Daily Papers

We present LFM2, a family of Liquid Foundation Models designed for efficient on-device deployment and strong task capabilities. Using hardware-in-the-loop architecture search under edge latency and memory constraints, we obtain a compact hybrid backbone that combines gated short convolutions with a small number of grouped query attention blocks, delivering up to 2x faster prefill and decode on CPUs compared to similarly sized models. The LFM2 family covers 350M-8.3B parameters, including dense models (350M, 700M, 1.2B, 2.6B) and a mixture-of-experts variant (8.3B total, 1.5B active), all with 32K context length. LFM2's training pipeline includes a tempered, decoupled Top-K knowledge distillation objective that avoids support mismatch; curriculum learning with difficulty-ordered data; and a three-stage post-training recipe of supervised fine-tuning, length-normalized preference optimization, and model merging. Pre-trained on 10-12T tokens, LFM2 models achieve strong results across diverse benchmarks; for example, LFM2-2.6B reaches 79.56% on IFEval and 82.41% on GSM8K. We further build multimodal and retrieval variants: LFM2-VL for vision-language tasks, LFM2-Audio for speech, and LFM2-ColBERT for retrieval. LFM2-VL supports tunable accuracy-latency tradeoffs via token-efficient visual processing, while LFM2-Audio separates audio input and output pathways to enable real-time speech-to-speech interaction competitive with models 3x larger. LFM2-ColBERT provides a low-latency encoder for queries and documents, enabling high-performance retrieval across multiple languages. All models are released with open weights and deployment packages for ExecuTorch, llama.cpp, and vLLM, making LFM2 a practical base for edge applications that need fast, memory-efficient inference and strong task capabilities.

Open-vocabulary detectors achieve impressive performance on COCO, but often fail to generalize to real-world datasets with out-of-distribution classes not typically found in their pre-training. Rather than simply fine-tuning a heavy-weight vision-language model (VLM) for new domains, we introduce RF-DETR, a light-weight specialist detection transformer that discovers accuracy-latency Pareto curves for any target dataset with weight-sharing neural architecture search (NAS). Our approach fine-tunes a pre-trained base network on a target dataset and evaluates thousands of network configurations with different accuracy-latency tradeoffs without re-training. Further, we revisit the "tunable knobs" for NAS to improve the transferability of DETRs to diverse target domains. Notably, RF-DETR significantly improves on prior state-of-the-art real-time methods on COCO and Roboflow100-VL. RF-DETR (nano) achieves 48.0 AP on COCO, beating D-FINE (nano) by 5.3 AP at similar latency, and RF-DETR (2x-large) outperforms GroundingDINO (tiny) by 1.2 AP on Roboflow100-VL while running 20x as fast. To the best of our knowledge, RF-DETR (2x-large) is the first real-time detector to surpass 60 AP on COCO. Our code is at https://github.com/roboflow/rf-detr

Enterprise chatbots, powered by generative AI, are emerging as key applications to enhance employee productivity. Retrieval Augmented Generation (RAG), Large Language Models (LLMs), and orchestration frameworks like Langchain and Llamaindex are crucial for building these chatbots. However, creating effective enterprise chatbots is challenging and requires meticulous RAG pipeline engineering. This includes fine-tuning embeddings and LLMs, extracting documents from vector databases, rephrasing queries, reranking results, designing prompts, honoring document access controls, providing concise responses, including references, safeguarding personal information, and building orchestration agents. We present a framework for building RAG-based chatbots based on our experience with three NVIDIA chatbots: for IT/HR benefits, financial earnings, and general content. Our contributions are three-fold: introducing the FACTS framework (Freshness, Architectures, Cost, Testing, Security), presenting fifteen RAG pipeline control points, and providing empirical results on accuracy-latency tradeoffs between large and small LLMs. To the best of our knowledge, this is the first paper of its kind that provides a holistic view of the factors as well as solutions for building secure enterprise-grade chatbots."

FPGAs provide a flexible and efficient platform to accelerate rapidly-changing algorithms for computer vision. The majority of existing work focuses on accelerating image classification, while other fundamental vision problems, including object detection and instance segmentation, have not been adequately addressed. Compared with image classification, detection problems are more sensitive to the spatial variance of objects, and therefore, require specialized convolutions to aggregate spatial information. To address this, recent work proposes dynamic deformable convolution to augment regular convolutions. Regular convolutions process a fixed grid of pixels across all the spatial locations in an image, while dynamic deformable convolutions may access arbitrary pixels in the image and the access pattern is input-dependent and varies per spatial location. These properties lead to inefficient memory accesses of inputs with existing hardware. In this work, we first investigate the overhead of the deformable convolution on embedded FPGA SoCs, and then show the accuracy-latency tradeoffs for a set of algorithm modifications including full versus depthwise, fixed-shape, and limited-range. These modifications benefit the energy efficiency for embedded devices in general as they reduce the compute complexity. We then build an efficient object detection network with modified deformable convolutions and quantize the network using state-of-the-art quantization methods. We implement a unified hardware engine on FPGA to support all the operations in the network. Preliminary experiments show that little accuracy is compromised and speedup can be achieved with our co-design optimization for the deformable convolution.

Multimodal Large Language Models (MLLMs) have shown impressive capabilities in reasoning, yet come with substantial computational cost, limiting their deployment in resource-constrained settings. Despite recent efforts on improving the efficiency of MLLMs, prior solutions fall short in responding to varying runtime conditions, in particular changing resource availability (e.g., contention due to the execution of other programs on the device). To bridge this gap, we introduce AdaLLaVA, an adaptive inference framework that learns to dynamically reconfigure operations in an MLLM during inference, accounting for the input data and a latency budget. We conduct extensive experiments across benchmarks involving question-answering, reasoning, and hallucination. Our results show that AdaLLaVA effectively adheres to input latency budget, achieving varying accuracy and latency tradeoffs at runtime. Further, we demonstrate that AdaLLaVA adapts to both input latency and content, can be integrated with token selection for enhanced efficiency, and generalizes across MLLMs. Our project webpage with code release is at https://zhuoyan-xu.github.io/ada-llava/.

Edge intelligence paradigm is increasingly demanded by the emerging autonomous systems, such as robotics. Beyond ensuring privacy-preserving operation and resilience in connectivity-limited environments, edge deployment offers significant energy and cost advantages over cloud-based solutions. However, deploying large language models (LLMs) for reasoning tasks on edge GPUs faces critical challenges from strict latency constraints and limited computational resources. To navigate these constraints, developers must balance multiple design factors - choosing reasoning versus non-reasoning architectures, selecting appropriate model sizes, allocating token budgets, and applying test-time scaling strategies - to meet target latency and optimize accuracy. Yet guidance on optimal combinations of these variables remains scarce. In this work, we present EdgeReasoning, a comprehensive study characterizing the deployment of reasoning LLMs on edge GPUs. We systematically quantify latency-accuracy tradeoffs across various LLM architectures and model sizes. We systematically evaluate prompt-based and model-tuning-based techniques for reducing reasoning token length while maintaining performance quality. We further profile test-time scaling methods with varying degrees of parallelism to maximize accuracy under strict latency budgets. Through these analyses, EdgeReasoning maps the Pareto frontier of achievable accuracy-latency configurations, offering systematic guidance for optimal edge deployment of reasoning LLMs.

Contrastive pretraining of image-text foundation models, such as CLIP, demonstrated excellent zero-shot performance and improved robustness on a wide range of downstream tasks. However, these models utilize large transformer-based encoders with significant memory and latency overhead which pose challenges for deployment on mobile devices. In this work, we introduce MobileCLIP -- a new family of efficient image-text models optimized for runtime performance along with a novel and efficient training approach, namely multi-modal reinforced training. The proposed training approach leverages knowledge transfer from an image captioning model and an ensemble of strong CLIP encoders to improve the accuracy of efficient models. Our approach avoids train-time compute overhead by storing the additional knowledge in a reinforced dataset. MobileCLIP sets a new state-of-the-art latency-accuracy tradeoff for zero-shot classification and retrieval tasks on several datasets. Our MobileCLIP-S2 variant is 2.3times faster while more accurate compared to previous best CLIP model based on ViT-B/16. We further demonstrate the effectiveness of our multi-modal reinforced training by training a CLIP model based on ViT-B/16 image backbone and achieving +2.9% average performance improvement on 38 evaluation benchmarks compared to the previous best. Moreover, we show that the proposed approach achieves 10times-1000times improved learning efficiency when compared with non-reinforced CLIP training.

In this work we explore the latency and accuracy of keyword spotting (KWS) models in streaming and non-streaming modes on mobile phones. NN model conversion from non-streaming mode (model receives the whole input sequence and then returns the classification result) to streaming mode (model receives portion of the input sequence and classifies it incrementally) may require manual model rewriting. We address this by designing a Tensorflow/Keras based library which allows automatic conversion of non-streaming models to streaming ones with minimum effort. With this library we benchmark multiple KWS models in both streaming and non-streaming modes on mobile phones and demonstrate different tradeoffs between latency and accuracy. We also explore novel KWS models with multi-head attention which reduce the classification error over the state-of-art by 10% on Google speech commands data sets V2. The streaming library with all experiments is open-sourced.

Retrieval-Augmented Generation (RAG) pipelines are central to applying large language models (LLMs) to proprietary or dynamic data. However, building effective RAG flows is complex, requiring careful selection among vector databases, embedding models, text splitters, retrievers, and synthesizing LLMs. The challenge deepens with the rise of agentic paradigms. Modules like verifiers, rewriters, and rerankers-each with intricate hyperparameter dependencies have to be carefully tuned. Balancing tradeoffs between latency, accuracy, and cost becomes increasingly difficult in performance-sensitive applications. We introduce syftr, a framework that performs efficient multi-objective search over a broad space of agentic and non-agentic RAG configurations. Using Bayesian Optimization, syftr discovers Pareto-optimal flows that jointly optimize task accuracy and cost. A novel early-stopping mechanism further improves efficiency by pruning clearly suboptimal candidates. Across multiple RAG benchmarks, syftr finds flows which are on average approximately 9 times cheaper while preserving most of the accuracy of the most accurate flows on the Pareto-frontier. Furthermore, syftr's ability to design and optimize allows integrating new modules, making it even easier and faster to realize high-performing generative AI pipelines.

Scene Graph Generation (SGG) is a task that encodes visual relationships between objects in images as graph structures. SGG shows significant promise as a foundational component for downstream tasks, such as reasoning for embodied agents. To enable real-time applications, SGG must address the trade-off between performance and inference speed. However, current methods tend to focus on one of the following: (1) improving relation prediction accuracy, (2) enhancing object detection accuracy, or (3) reducing latency, without aiming to balance all three objectives simultaneously. To address this limitation, we propose the Real-time Efficiency and Accuracy Compromise for Tradeoffs in Scene Graph Generation (REACT) architecture, which achieves the highest inference speed among existing SGG models, improving object detection accuracy without sacrificing relation prediction performance. Compared to state-of-the-art approaches, REACT is 2.7 times faster and improves object detection accuracy by 58\%. Furthermore, our proposal significantly reduces model size, with an average of 5.5x fewer parameters. The code is available at https://github.com/Maelic/SGG-Benchmark

Fluid voice-to-voice interaction requires reliable and low-latency detection of when a user has finished speaking. Traditional audio-silence end-pointers add hundreds of milliseconds of delay and fail under hesitations or language-specific phenomena. We present, to our knowledge, the first systematic study of Thai text-only end-of-turn (EOT) detection for real-time agents. We compare zero-shot and few-shot prompting of compact LLMs to supervised fine-tuning of lightweight transformers. Using transcribed subtitles from the YODAS corpus and Thai-specific linguistic cues (e.g., sentence-final particles), we formulate EOT as a binary decision over token boundaries. We report a clear accuracy-latency tradeoff and provide a public-ready implementation plan. This work establishes a Thai baseline and demonstrates that small, fine-tuned models can deliver near-instant EOT decisions suitable for on-device agents.

VIP navigation requires multiple DNN models for identification, posture analysis, and depth estimation to ensure safe mobility. Using a hazard vest as a unique identifier enhances visibility while selecting the right DNN model and computing device balances accuracy and real-time performance. We present Ocularone-Bench, which is a benchmark suite designed to address the lack of curated datasets for uniquely identifying individuals in crowded environments and the need for benchmarking DNN inference times on resource-constrained edge devices. The suite evaluates the accuracy-latency trade-offs of YOLO models retrained on this dataset and benchmarks inference times of situation awareness models across edge accelerators and high-end GPU workstations. Our study on NVIDIA Jetson devices and RTX 4090 workstation demonstrates significant improvements in detection accuracy, achieving up to 99.4% precision, while also providing insights into real-time feasibility for mobile deployment. Beyond VIP navigation, Ocularone-Bench is applicable to senior citizens, children and worker safety monitoring, and other vision-based applications.

The rising popularity of intelligent mobile devices and the daunting computational cost of deep learning-based models call for efficient and accurate on-device inference schemes. We propose a quantization scheme that allows inference to be carried out using integer-only arithmetic, which can be implemented more efficiently than floating point inference on commonly available integer-only hardware. We also co-design a training procedure to preserve end-to-end model accuracy post quantization. As a result, the proposed quantization scheme improves the tradeoff between accuracy and on-device latency. The improvements are significant even on MobileNets, a model family known for run-time efficiency, and are demonstrated in ImageNet classification and COCO detection on popular CPUs.

We present YOLOBench, a benchmark comprised of 550+ YOLO-based object detection models on 4 different datasets and 4 different embedded hardware platforms (x86 CPU, ARM CPU, Nvidia GPU, NPU). We collect accuracy and latency numbers for a variety of YOLO-based one-stage detectors at different model scales by performing a fair, controlled comparison of these detectors with a fixed training environment (code and training hyperparameters). Pareto-optimality analysis of the collected data reveals that, if modern detection heads and training techniques are incorporated into the learning process, multiple architectures of the YOLO series achieve a good accuracy-latency trade-off, including older models like YOLOv3 and YOLOv4. We also evaluate training-free accuracy estimators used in neural architecture search on YOLOBench and demonstrate that, while most state-of-the-art zero-cost accuracy estimators are outperformed by a simple baseline like MAC count, some of them can be effectively used to predict Pareto-optimal detection models. We showcase that by using a zero-cost proxy to identify a YOLO architecture competitive against a state-of-the-art YOLOv8 model on a Raspberry Pi 4 CPU. The code and data are available at https://github.com/Deeplite/deeplite-torch-zoo

While neural network hardware accelerators provide a substantial amount of raw compute throughput, the models deployed on them must be co-designed for the underlying hardware architecture to obtain the optimal system performance. We present a class of computer vision models designed using hardware-aware neural architecture search and customized to run on the Edge TPU, Google's neural network hardware accelerator for low-power, edge devices. For the Edge TPU in Coral devices, these models enable real-time image classification performance while achieving accuracy typically seen only with larger, compute-heavy models running in data centers. On Pixel 4's Edge TPU, these models improve the accuracy-latency tradeoff over existing SoTA mobile models.

Scaling laws are powerful tools to predict the performance of large language models. However, current scaling laws fall short of accounting for inference costs. In this work, we first show that model architecture affects inference latency, where models of the same size can have up to 3.5x difference in latency. To tackle this challenge, we modify the Chinchilla scaling laws to co-optimize the model parameter count, the number of training tokens, and the model architecture. Due to the reason that models of similar training loss exhibit gaps in downstream evaluation, we also propose a novel method to train inference-efficient models based on the revised scaling laws. We perform extensive empirical studies to fit and evaluate our inference-aware scaling laws. We vary model parameters from 80M to 1B, training tokens from 1.6B to 30B, and model shapes, training a total of 63 models. Guided by our inference-efficient scaling law and model selection method, we release the Morph-1B model, which improves inference latency by 1.8x while maintaining accuracy on downstream tasks compared to open-source models, pushing the Pareto frontier of accuracy-latency tradeoff.

Deep Neural Networks (DNNs) have been a large driver and enabler for AI breakthroughs in recent years. These models have been getting larger in their attempt to become more accurate and tackle new upcoming use-cases, including AR/VR and intelligent assistants. However, the training process of such large models is a costly and time-consuming process, which typically yields a single model to fit all targets. To mitigate this, various techniques have been proposed in the literature, including pruning, sparsification or quantization of the model weights and updates. While able to achieve high compression rates, they often incur computational overheads or accuracy penalties. Alternatively, factorization methods have been leveraged to incorporate low-rank compression in the training process. Similarly, such techniques (e.g.,~SVD) frequently rely on the computationally expensive decomposition of layers and are potentially sub-optimal for non-linear models, such as DNNs. In this work, we take a further step in designing efficient low-rank models and propose Maestro, a framework for trainable low-rank layers. Instead of regularly applying a priori decompositions such as SVD, the low-rank structure is built into the training process through a generalized variant of Ordered Dropout. This method imposes an importance ordering via sampling on the decomposed DNN structure. Our theoretical analysis demonstrates that our method recovers the SVD decomposition of linear mapping on uniformly distributed data and PCA for linear autoencoders. We further apply our technique on DNNs and empirically illustrate that Maestro enables the extraction of lower footprint models that preserve model performance while allowing for graceful accuracy-latency tradeoff for the deployment to devices of different capabilities.

In this work, we present QuickSRNet, an efficient super-resolution architecture for real-time applications on mobile platforms. Super-resolution clarifies, sharpens, and upscales an image to higher resolution. Applications such as gaming and video playback along with the ever-improving display capabilities of TVs, smartphones, and VR headsets are driving the need for efficient upscaling solutions. While existing deep learning-based super-resolution approaches achieve impressive results in terms of visual quality, enabling real-time DL-based super-resolution on mobile devices with compute, thermal, and power constraints is challenging. To address these challenges, we propose QuickSRNet, a simple yet effective architecture that provides better accuracy-to-latency trade-offs than existing neural architectures for single-image super resolution. We present training tricks to speed up existing residual-based super-resolution architectures while maintaining robustness to quantization. Our proposed architecture produces 1080p outputs via 2x upscaling in 2.2 ms on a modern smartphone, making it ideal for high-fps real-time applications.

The transformer model has gained widespread adoption in computer vision tasks in recent times. However, due to the quadratic time and memory complexity of self-attention, which is proportional to the number of input tokens, most existing Vision Transformers (ViTs) encounter challenges in achieving efficient performance in practical industrial deployment scenarios, such as TensorRT and CoreML, where traditional CNNs excel. Although some recent attempts have been made to design CNN-Transformer hybrid architectures to tackle this problem, their overall performance has not met expectations. To tackle these challenges, we propose an efficient hybrid ViT architecture named FMViT. This approach enhances the model's expressive power by blending high-frequency features and low-frequency features with varying frequencies, enabling it to capture both local and global information effectively. Additionally, we introduce deploy-friendly mechanisms such as Convolutional Multigroup Reparameterization (gMLP), Lightweight Multi-head Self-Attention (RLMHSA), and Convolutional Fusion Block (CFB) to further improve the model's performance and reduce computational overhead. Our experiments demonstrate that FMViT surpasses existing CNNs, ViTs, and CNNTransformer hybrid architectures in terms of latency/accuracy trade-offs for various vision tasks. On the TensorRT platform, FMViT outperforms Resnet101 by 2.5% (83.3% vs. 80.8%) in top-1 accuracy on the ImageNet dataset while maintaining similar inference latency. Moreover, FMViT achieves comparable performance with EfficientNet-B5, but with a 43% improvement in inference speed. On CoreML, FMViT outperforms MobileOne by 2.6% in top-1 accuracy on the ImageNet dataset, with inference latency comparable to MobileOne (78.5% vs. 75.9%). Our code can be found at https://github.com/tany0699/FMViT.

Recent work in speech-to-speech translation (S2ST) has focused primarily on offline settings, where the full input utterance is available before any output is given. This, however, is not reasonable in many real-world scenarios. In latency-sensitive applications, rather than waiting for the full utterance, translations should be spoken as soon as the information in the input is present. In this work, we introduce a system for simultaneous S2ST targeting real-world use cases. Our system supports translation from 57 languages to English with tunable parameters for dynamically adjusting the latency of the output -- including four policies for determining when to speak an output sequence. We show that these policies achieve offline-level accuracy with minimal increases in latency over a Greedy (wait-k) baseline. We open-source our evaluation code and interactive test script to aid future SimulS2ST research and application development.

Neural information retrieval (IR) systems have progressed rapidly in recent years, in large part due to the release of publicly available benchmarking tasks. Unfortunately, some dimensions of this progress are illusory: the majority of the popular IR benchmarks today focus exclusively on downstream task accuracy and thus conceal the costs incurred by systems that trade away efficiency for quality. Latency, hardware cost, and other efficiency considerations are paramount to the deployment of IR systems in user-facing settings. We propose that IR benchmarks structure their evaluation methodology to include not only metrics of accuracy, but also efficiency considerations such as a query latency and the corresponding cost budget for a reproducible hardware setting. For the popular IR benchmarks MS MARCO and XOR-TyDi, we show how the best choice of IR system varies according to how these efficiency considerations are chosen and weighed. We hope that future benchmarks will adopt these guidelines toward more holistic IR evaluation.

Differentiable neural architecture search methods became popular in recent years, mainly due to their low search costs and flexibility in designing the search space. However, these methods suffer the difficulty in optimizing network, so that the searched network is often unfriendly to hardware. This paper deals with this problem by adding a differentiable latency loss term into optimization, so that the search process can tradeoff between accuracy and latency with a balancing coefficient. The core of latency prediction is to encode each network architecture and feed it into a multi-layer regressor, with the training data which can be easily collected from randomly sampling a number of architectures and evaluating them on the hardware. We evaluate our approach on NVIDIA Tesla-P100 GPUs. With 100K sampled architectures (requiring a few hours), the latency prediction module arrives at a relative error of lower than 10%. Equipped with this module, the search method can reduce the latency by 20% meanwhile preserving the accuracy. Our approach also enjoys the ability of being transplanted to a wide range of hardware platforms with very few efforts, or being used to optimizing other non-differentiable factors such as power consumption.

Weight-only quantization is widely used to mitigate the memory-bound nature of LLM inference. Codebook-based methods extend this trend by achieving strong accuracy in the extremely low-bit regime (e.g., 2-bit). However, current kernels rely on dequantization, which repeatedly fetches centroids and reconstructs weights, incurring substantial latency and cache pressure. We present CodeGEMM, a codebook-centric GEMM kernel that replaces dequantization with precomputed inner products between centroids and activations stored in a lightweight Psumbook. At inference, code indices directly gather these partial sums, eliminating per-element lookups and reducing the on-chip footprint. The kernel supports the systematic exploration of latency-memory-accuracy trade-offs under a unified implementation. On Llama-3 models, CodeGEMM delivers 1.83x (8B) and 8.93x (70B) speedups in the 2-bit configuration compared to state-of-the-art codebook-based quantization at comparable accuracy and further improves computing efficiency and memory subsystem utilization.

Test-Time Scaling (TTS) has proven effective in improving the performance of Large Language Models (LLMs) during inference. However, existing research has overlooked the efficiency of TTS from a latency-sensitive perspective. Through a latency-aware evaluation of representative TTS methods, we demonstrate that a compute-optimal TTS does not always result in the lowest latency in scenarios where latency is critical. To address this gap and achieve latency-optimal TTS, we propose two key approaches by optimizing the concurrency configurations: (1) branch-wise parallelism, which leverages multiple concurrent inference branches, and (2) sequence-wise parallelism, enabled by speculative decoding. By integrating these two approaches and allocating computational resources properly to each, our latency-optimal TTS enables a 32B model to reach 82.3% accuracy on MATH-500 within 1 minute and a smaller 3B model to achieve 72.4% within 10 seconds. Our work emphasizes the importance of latency-aware TTS and demonstrates its ability to deliver both speed and accuracy in latency-sensitive scenarios.

Efficient deployment of small language models (SLMs) is essential for numerous real-world applications with stringent latency constraints. While previous work on SLM design has primarily focused on reducing the number of parameters to achieve parameter-optimal SLMs, parameter efficiency does not necessarily translate into proportional real-device speed-ups. This work aims to identify the key determinants of SLMs' real-device latency and offer generalizable principles and methodologies for SLM design and training when real-device latency is the primary consideration. Specifically, we identify two central architectural factors: depth-width ratios and operator choices. The former is crucial for small-batch-size latency, while the latter affects both latency and large-batch-size throughput. In light of this, we first study latency-optimal depth-width ratios, with the key finding that although deep-thin models generally achieve better accuracy under the same parameter budget, they may not lie on the accuracy-latency trade-off frontier. Next, we explore emerging efficient attention alternatives to evaluate their potential as candidate building operators. Using the identified promising operators, we construct an evolutionary search framework to automatically discover latency-optimal combinations of these operators within hybrid SLMs, thereby advancing the accuracy-latency frontier. In addition to architectural improvements, we further enhance SLM training using a weight normalization technique that enables more effective weight updates and improves final convergence. Combining these methods, we introduce a new family of hybrid SLMs, called Nemotron-Flash, which significantly advances the accuracy-efficiency frontier of state-of-the-art SLMs, e.g., achieving over +5.5% average accuracy, 1.3x/1.9x lower latency, and 18.7x/45.6x higher throughput compared to Qwen3-1.7B/0.6B, respectively.

Large Language Model (LLM)-based search agents have shown remarkable capabilities in solving complex tasks by dynamically decomposing problems and addressing them through interleaved reasoning and retrieval. However, this interleaved paradigm introduces substantial efficiency bottlenecks. First, we observe that both highly accurate and overly approximate retrieval methods degrade system efficiency: exact search incurs significant retrieval overhead, while coarse retrieval requires additional reasoning steps during generation. Second, we identify inefficiencies in system design, including improper scheduling and frequent retrieval stalls, which lead to cascading latency -- where even minor delays in retrieval amplify end-to-end inference time. To address these challenges, we introduce SearchAgent-X, a high-efficiency inference framework for LLM-based search agents. SearchAgent-X leverages high-recall approximate retrieval and incorporates two key techniques: priority-aware scheduling and non-stall retrieval. Extensive experiments demonstrate that SearchAgent-X consistently outperforms state-of-the-art systems such as vLLM and HNSW-based retrieval across diverse tasks, achieving up to 3.4times higher throughput and 5times lower latency, without compromising generation quality. SearchAgent-X is available at https://github.com/tiannuo-yang/SearchAgent-X.

AI agents -- systems that combine foundation models with reasoning, planning, memory, and tool use -- are rapidly becoming a practical interface between natural-language intent and real-world computation. This survey synthesizes the emerging landscape of AI agent architectures across: (i) deliberation and reasoning (e.g., chain-of-thought-style decomposition, self-reflection and verification, and constraint-aware decision making), (ii) planning and control (from reactive policies to hierarchical and multi-step planners), and (iii) tool calling and environment interaction (retrieval, code execution, APIs, and multimodal perception). We organize prior work into a unified taxonomy spanning agent components (policy/LLM core, memory, world models, planners, tool routers, and critics), orchestration patterns (single-agent vs.\ multi-agent; centralized vs.\ decentralized coordination), and deployment settings (offline analysis vs.\ online interactive assistance; safety-critical vs.\ open-ended tasks). We discuss key design trade-offs -- latency vs.\ accuracy, autonomy vs.\ controllability, and capability vs.\ reliability -- and highlight how evaluation is complicated by non-determinism, long-horizon credit assignment, tool and environment variability, and hidden costs such as retries and context growth. Finally, we summarize measurement and benchmarking practices (task suites, human preference and utility metrics, success under constraints, robustness and security) and identify open challenges including verification and guardrails for tool actions, scalable memory and context management, interpretability of agent decisions, and reproducible evaluation under realistic workloads.

Reasoning-based language models have demonstrated strong performance across various domains, with the most notable gains seen in mathematical and coding tasks. Recent research has shown that reasoning also offers significant benefits for LLM safety and guardrail applications. In this work, we conduct a comprehensive analysis of training reasoning-based guardrail models for content moderation, with an emphasis on generalization to custom safety policies at inference time. Our study focuses on two key dimensions: data efficiency and inference efficiency. On the data front, we find that reasoning-based models exhibit strong sample efficiency, achieving competitive performance with significantly fewer training examples than their non-reasoning counterparts. This unlocks the potential to repurpose the remaining data for mining high-value, difficult samples that further enhance model performance. On the inference side, we evaluate practical trade-offs by introducing reasoning budgets, examining the impact of reasoning length on latency and accuracy, and exploring dual-mode training to allow runtime control over reasoning behavior. Our findings will provide practical insights for researchers and developers to effectively and efficiently train and deploy reasoning-based guardrails models in real-world systems.

Transformer models are deployed in a wide range of settings, from multi-accelerator clusters to standalone mobile phones. The diverse inference constraints in these scenarios necessitate practitioners to train foundation models such as PaLM 2, Llama, & ViTs as a series of models of varying sizes. Due to significant training costs, only a select few model sizes are trained and supported, limiting more fine-grained control over relevant tradeoffs, including latency, cost, and accuracy. This work introduces MatFormer, a nested Transformer architecture designed to offer elasticity in a variety of deployment constraints. Each Feed Forward Network (FFN) block of a MatFormer model is jointly optimized with a few nested smaller FFN blocks. This training procedure allows for the Mix'n'Match of model granularities across layers -- i.e., a trained universal MatFormer model enables extraction of hundreds of accurate smaller models, which were never explicitly optimized. We empirically demonstrate MatFormer's effectiveness across different model classes (decoders & encoders), modalities (language & vision), and scales (up to 2.6B parameters). We find that a 2.6B decoder-only MatFormer language model (MatLM) allows us to extract smaller models spanning from 1.5B to 2.6B, each exhibiting comparable validation loss and one-shot downstream evaluations to their independently trained counterparts. Furthermore, we observe that smaller encoders extracted from a universal MatFormer-based ViT (MatViT) encoder preserve the metric-space structure for adaptive large-scale retrieval. Finally, we showcase that speculative decoding with the accurate and consistent submodels extracted from MatFormer can further reduce inference latency.

For deployment, neural architecture search should be hardware-aware, in order to satisfy the device-specific constraints (e.g., memory usage, latency and energy consumption) and enhance the model efficiency. Existing methods on hardware-aware NAS collect a large number of samples (e.g., accuracy and latency) from a target device, either builds a lookup table or a latency estimator. However, such approach is impractical in real-world scenarios as there exist numerous devices with different hardware specifications, and collecting samples from such a large number of devices will require prohibitive computational and monetary cost. To overcome such limitations, we propose Hardware-adaptive Efficient Latency Predictor (HELP), which formulates the device-specific latency estimation problem as a meta-learning problem, such that we can estimate the latency of a model's performance for a given task on an unseen device with a few samples. To this end, we introduce novel hardware embeddings to embed any devices considering them as black-box functions that output latencies, and meta-learn the hardware-adaptive latency predictor in a device-dependent manner, using the hardware embeddings. We validate the proposed HELP for its latency estimation performance on unseen platforms, on which it achieves high estimation performance with as few as 10 measurement samples, outperforming all relevant baselines. We also validate end-to-end NAS frameworks using HELP against ones without it, and show that it largely reduces the total time cost of the base NAS method, in latency-constrained settings. Code is available at https://github.com/HayeonLee/HELP.

The large number of parameters in Pretrained Language Models enhance their performance, but also make them resource-intensive, making it challenging to deploy them on commodity hardware like a single GPU. Due to the memory and power limitations of these devices, model compression techniques are often used to decrease both the model's size and its inference latency. This usually results in a trade-off between model accuracy and efficiency. Therefore, optimizing this balance is essential for effectively deploying LLMs on commodity hardware. A significant portion of the efficiency challenge is the Feed-forward network (FFN) component, which accounts for roughly 2{3} total parameters and inference latency. In this paper, we first observe that only a few neurons of FFN module have large output norm for any input tokens, a.k.a. heavy hitters, while the others are sparsely triggered by different tokens. Based on this observation, we explicitly split the FFN into two parts according to the heavy hitters. We improve the efficiency-accuracy trade-off of existing compression methods by allocating more resource to FFN parts with heavy hitters. In practice, our method can reduce model size by 43.1\% and bring 1.25sim1.56times wall clock time speedup on different hardware with negligible accuracy drop.

Large language models (LLMs) power many state-of-the-art systems in natural language processing. However, these models are extremely computationally expensive, even at inference time, raising the natural question: when is the extra cost of deploying a larger model worth the anticipated boost in capabilities? Better understanding this tradeoff fundamentally could benefit from an inference efficiency metric that is both (i) easily comparable across models from different providers, and (ii) representative of the true cost of running queries in an isolated performance environment. Unfortunately, access to LLMs today is largely restricted to black-box text generation APIs and raw runtimes measured through this interface do not satisfy these desiderata: model providers can apply various software and hardware optimizations orthogonal to the model, and models served on shared infrastructure are susceptible to performance contention. To circumvent these problems, we propose a new metric for comparing inference efficiency across models. This metric puts models on equal footing as though they were served (i) on uniform hardware and software, and (ii) without performance contention. We call this metric the idealized runtime, and we propose a methodology to efficiently estimate this metric for autoregressive Transformer models. We also propose cost-aware variants that incorporate the number of accelerators needed to serve the model. Using these metrics, we compare ten state-of-the-art LLMs to provide the first analysis of inference efficiency-capability tradeoffs; we make several observations from this analysis, including the fact that the superior inference runtime performance of certain APIs is often a byproduct of optimizations within the API rather than the underlying model. Our methodology also facilitates the efficient comparison of different software and hardware stacks.

The sparse Mixture-of-Experts (MoE) architecture is increasingly favored for scaling Large Language Models (LLMs) efficiently, but it depends on heterogeneous compute and memory resources. These factors jointly affect system Cost, Accuracy, and Performance (CAP), making trade-offs inevitable. Existing benchmarks often fail to capture these trade-offs accurately, complicating practical deployment decisions. To address this, we introduce MoE-CAP, a benchmark specifically designed for MoE systems. Our analysis reveals that achieving an optimal balance across CAP is difficult with current hardware; MoE systems typically optimize two of the three dimensions at the expense of the third-a dynamic we term the MoE-CAP trade-off. To visualize this, we propose the CAP Radar Diagram. We further introduce sparsity-aware performance metrics-Sparse Memory Bandwidth Utilization (S-MBU) and Sparse Model FLOPS Utilization (S-MFU)-to enable accurate performance benchmarking of MoE systems across diverse hardware platforms and deployment scenarios.

Large Language Models (LLMs), such as OpenAI-o1 and DeepSeek-R1, have demonstrated strong reasoning capabilities. To further enhance LLM capabilities, recent agentic systems, such as Deep Research, incorporate web interactions into LLM reasoning to mitigate uncertainties and reduce potential errors. However, existing research predominantly focuses on reasoning performance, often neglecting the efficiency of agentic systems. In this work, we present a comprehensive empirical study that identifies efficiency bottlenecks in web-interactive agentic systems. We decompose end-to-end latency into two primary components: LLM API latency and web environment latency. We conduct a comprehensive empirical study across 15 models and 5 providers to demonstrate high variability in API-based agentic systems. We observe that web environment latency can contribute as much as 53.7% to the overall latency in a web-based agentic system. To improve latency, we propose SpecCache, a caching framework augmented with speculative execution that can reduce web environment overhead. Extensive evaluations on two standard benchmarks show that our approach improves the cache hit rate by up to 58x compared to a random caching strategy, while reducing web environment overhead by up to 3.2x, without degrading agentic system performance.

The deployment of deep neural networks (DNNs) on resource-constrained edge devices is frequently hindered by their significant computational and memory requirements. While partitioning and distributing a DNN across multiple devices is a well-established strategy to mitigate this challenge, prior research has largely focused on single-objective optimization, such as minimizing latency or maximizing throughput. This paper challenges that view by reframing DNN partitioning as a multi-objective optimization problem. We argue that in real-world scenarios, a complex trade-off between latency and throughput exists, which is further complicated by network variability. To address this, we introduce ParetoPipe, an open-source framework that leverages Pareto front analysis to systematically identify optimal partitioning strategies that balance these competing objectives. Our contributions are threefold: we benchmark pipeline partitioned inference on a heterogeneous testbed of Raspberry Pis and a GPU-equipped edge server; we identify Pareto-optimal points to analyze the latency-throughput trade-off under varying network conditions; and we release a flexible, open-source framework to facilitate distributed inference and benchmarking. This toolchain features dual communication backends, PyTorch RPC and a custom lightweight implementation, to minimize overhead and support broad experimentation.

Designing convolutional neural networks (CNN) for mobile devices is challenging because mobile models need to be small and fast, yet still accurate. Although significant efforts have been dedicated to design and improve mobile CNNs on all dimensions, it is very difficult to manually balance these trade-offs when there are so many architectural possibilities to consider. In this paper, we propose an automated mobile neural architecture search (MNAS) approach, which explicitly incorporate model latency into the main objective so that the search can identify a model that achieves a good trade-off between accuracy and latency. Unlike previous work, where latency is considered via another, often inaccurate proxy (e.g., FLOPS), our approach directly measures real-world inference latency by executing the model on mobile phones. To further strike the right balance between flexibility and search space size, we propose a novel factorized hierarchical search space that encourages layer diversity throughout the network. Experimental results show that our approach consistently outperforms state-of-the-art mobile CNN models across multiple vision tasks. On the ImageNet classification task, our MnasNet achieves 75.2% top-1 accuracy with 78ms latency on a Pixel phone, which is 1.8x faster than MobileNetV2 [29] with 0.5% higher accuracy and 2.3x faster than NASNet [36] with 1.2% higher accuracy. Our MnasNet also achieves better mAP quality than MobileNets for COCO object detection. Code is at https://github.com/tensorflow/tpu/tree/master/models/official/mnasnet

Large Language Models (LLMs) and other large foundation models have achieved noteworthy success, but their size exacerbates existing resource consumption and latency challenges. In particular, the large-scale deployment of these models is hindered by the significant resource requirements during inference. In this paper, we study two approaches for mitigating these challenges: employing a cache to store previous queries and learning a model multiplexer to choose from an ensemble of models for query processing. Theoretically, we provide an optimal algorithm for jointly optimizing both approaches to reduce the inference cost in both offline and online tabular settings. By combining a caching algorithm, namely Greedy Dual Size with Frequency (GDSF) or Least Expected Cost (LEC), with a model multiplexer, we achieve optimal rates in both offline and online settings. Empirically, simulations show that the combination of our caching and model multiplexing algorithms greatly improves over the baselines, with up to 50times improvement over the baseline when the ratio between the maximum cost and minimum cost is 100. Experiments on real datasets show a 4.3times improvement in FLOPs over the baseline when the ratio for FLOPs is 10, and a 1.8times improvement in latency when the ratio for average latency is 1.85.

Machine learning (ML) inference platforms are tasked with balancing two competing goals: ensuring high throughput given many requests, and delivering low-latency responses to support interactive applications. Unfortunately, existing platform knobs (e.g., batch sizes) fail to ease this fundamental tension, and instead only enable users to harshly trade off one property for the other. This paper explores an alternate strategy to taming throughput-latency tradeoffs by changing the granularity at which inference is performed. We present Apparate, a system that automatically applies and manages early exits (EEs) in ML models, whereby certain inputs can exit with results at intermediate layers. To cope with the time-varying overhead and accuracy challenges that EEs bring, Apparate repurposes exits to provide continual feedback that powers several novel runtime monitoring and adaptation strategies. Apparate lowers median response latencies by 40.5--91.5% and 10.0--24.2% for diverse CV and NLP classification workloads, and median time-per-token latencies by 22.6--77.9% for generative scenarios, without affecting throughputs or violating tight accuracy constraints.

The application of on-device language models (ODLMs) on resource-constrained edge devices is a multi-dimensional problem that strikes a fine balance between computational effectiveness, memory, power usage, and linguistic capacity across heterogeneous tasks. This holistic study conducts a thorough investigation of the trade-offs between domain-specific optimization and cross-domain robustness, culminating in the proposal of the Generalized Edge Model (GEM), a new architecture that aims to balance specialization and generalization in a harmonious manner. With a rigorous experimental approach testing 47 well-chosen benchmarks in eight domains--healthcare, law, finance, STEM, commonsense, conversational AI, multilingual, and domain-adaptive tasks--we show that conventional optimization techniques decrease target task perplexity by 18-25% but result in a precipitous decline in general-task performance with F1 scores decreasing by 12-29%, as reported by Liu et al. GEM employs a Sparse Cross-Attention Router (SCAR) to dynamically allocate computation to a variable number of computing resources with a cross-domain F1 accuracy of 0.89 on less than 100ms latency across Raspberry Pi 4, Pixel 6, iPhone 13, and bespoke custom neural processing units (NPUs). Compared to GPT-4 Lite, GEM enhances the general-task level by 7% with respect and parity in domain-specific performance. We propose three new measurement tools--Domain Specialization Index (DSI), Generalization Gap (GG), and Cross-Domain Transfer Ratio (CDTR)--which show strong correlation between model compression intensity and brittleness.

Mobile devices such as smartphones and autonomous vehicles increasingly rely on deep neural networks (DNNs) to execute complex inference tasks such as image classification and speech recognition, among others. However, continuously executing the entire DNN on mobile devices can quickly deplete their battery. Although task offloading to cloud/edge servers may decrease the mobile device's computational burden, erratic patterns in channel quality, network, and edge server load can lead to a significant delay in task execution. Recently, approaches based on split computing (SC) have been proposed, where the DNN is split into a head and a tail model, executed respectively on the mobile device and on the edge server. Ultimately, this may reduce bandwidth usage as well as energy consumption. Another approach, called early exiting (EE), trains models to embed multiple "exits" earlier in the architecture, each providing increasingly higher target accuracy. Therefore, the trade-off between accuracy and delay can be tuned according to the current conditions or application demands. In this paper, we provide a comprehensive survey of the state of the art in SC and EE strategies by presenting a comparison of the most relevant approaches. We conclude the paper by providing a set of compelling research challenges.

The success of transformers in computer vision has led to several attempts to adapt them for mobile devices, but their performance remains unsatisfactory in some real-world applications. To address this issue, we propose PP-MobileSeg, a semantic segmentation model that achieves state-of-the-art performance on mobile devices. PP-MobileSeg comprises three novel parts: the StrideFormer backbone, the Aggregated Attention Module (AAM), and the Valid Interpolate Module (VIM). The four-stage StrideFormer backbone is built with MV3 blocks and strided SEA attention, and it is able to extract rich semantic and detailed features with minimal parameter overhead. The AAM first filters the detailed features through semantic feature ensemble voting and then combines them with semantic features to enhance the semantic information. Furthermore, we proposed VIM to upsample the downsampled feature to the resolution of the input image. It significantly reduces model latency by only interpolating classes present in the final prediction, which is the most significant contributor to overall model latency. Extensive experiments show that PP-MobileSeg achieves a superior tradeoff between accuracy, model size, and latency compared to other methods. On the ADE20K dataset, PP-MobileSeg achieves 1.57% higher accuracy in mIoU than SeaFormer-Base with 32.9% fewer parameters and 42.3% faster acceleration on Qualcomm Snapdragon 855. Source codes are available at https://github.com/PaddlePaddle/PaddleSeg/tree/release/2.8.

Recent advancements in large language models (LLMs) have prompted interest in deploying these models on mobile devices to enable new applications without relying on cloud connectivity. However, the efficiency constraints of deploying LLMs on resource-limited devices present significant challenges. In this paper, we conduct a comprehensive measurement study to evaluate the efficiency tradeoffs between mobile-based, edge-based, and cloud-based deployments for LLM applications. We implement AutoLife-Lite, a simplified LLM-based application that analyzes smartphone sensor data to infer user location and activity contexts. Our experiments reveal that: (1) Only small-size LLMs (<4B parameters) can run successfully on powerful mobile devices, though they exhibit quality limitations compared to larger models; (2) Model compression is effective in lower the hardware requirement, but may lead to significant performance degradation; (3) The latency to run LLMs on mobile devices with meaningful output is significant (>30 seconds), while cloud services demonstrate better time efficiency (<10 seconds); (4) Edge deployments offer intermediate tradeoffs between latency and model capabilities, with different results on CPU-based and GPU-based settings. These findings provide valuable insights for system designers on the current limitations and future directions for on-device LLM applications.

Large Language Models (LLMs) are becoming integral to daily life, showcasing their vast potential across various Natural Language Processing (NLP) tasks. Beyond NLP, LLMs are increasingly used in software development tasks, such as code completion, modification, bug fixing, and code translation. Software engineers widely use tools like GitHub Copilot and Amazon Q, streamlining workflows and automating tasks with high accuracy. While the resource and energy intensity of LLM training is often highlighted, inference can be even more resource-intensive over time, as it's a continuous process with a high number of invocations. Therefore, developing resource-efficient alternatives for LLM inference is crucial for sustainability. This work proposes GREEN-CODE, a framework for energy-aware code generation in LLMs. GREEN-CODE performs dynamic early exit during LLM inference. We train a Reinforcement Learning (RL) agent that learns to balance the trade-offs between accuracy, latency, and energy consumption. Our approach is evaluated on two open-source LLMs, Llama 3.2 3B and OPT 2.7B, using the JavaCorpus and PY150 datasets. Results show that our method reduces the energy consumption between 23-50 % on average for code generation tasks without significantly affecting accuracy.

Speculative decoding accelerates LLM inference by using a draft model to look ahead, but gains are capped by the cost of autoregressive draft generation: increasing draft size elevates acceptance rates but introduces additional latency overhead exacerbating the speed-accuracy tradeoff. Prior methods (Medusa, Hydra, EAGLE) partially reduce draft cost but either degrade acceptance or introduce overheads that limit scaling. We present Mirror Speculative Decoding (Mirror-SD), an inference algorithm that breaks the latency-acceptance tradeoff. Mirror-SD launches branch-complete rollouts from early-exit signals in parallel with the target model's suffix and explicitly maps computation across heterogeneous accelerators (GPU and NPU) to exploit cross-device parallelism. The draft speculates forward continuations for the target to verify, while the target simultaneously speculates correction paths for the draft, converting speculation into two complementary execution pipelines. To further cut draft latency without weakening acceptance semantics, we add speculative streaming so the draft emits multiple tokens per step. This dual strategy of parallel heterogeneous execution plus multi-token speculative streaming pushes speculative decoding toward its ideal regime of high acceptance with low overhead. On SpecBench with server-scale models from 14B to 66B parameters, Mirror-SD delivers consistent end-to-end gains, achieving 2.8x-5.8x wall-time speedups across diverse tasks and a 30% average relative improvement over the strongest baseline, EAGLE3.

We study the problem of efficient generative inference for Transformer models, in one of its most challenging settings: large deep models, with tight latency targets and long sequence lengths. Better understanding of the engineering tradeoffs for inference for large Transformer-based models is important as use cases of these models are growing rapidly throughout application areas. We develop a simple analytical model for inference efficiency to select the best multi-dimensional partitioning techniques optimized for TPU v4 slices based on the application requirements. We combine these with a suite of low-level optimizations to achieve a new Pareto frontier on the latency and model FLOPS utilization (MFU) tradeoffs on 500B+ parameter models that outperforms the FasterTransformer suite of benchmarks. We further show that with appropriate partitioning, the lower memory requirements of multiquery attention (i.e. multiple query heads share single key/value head) enables scaling up to 32x larger context lengths. Finally, we achieve a low-batch-size latency of 29ms per token during generation (using int8 weight quantization) and a 76% MFU during large-batch-size processing of input tokens, while supporting a long 2048-token context length on the PaLM 540B parameter model.

Large language models (LLMs) have surged in popularity and are extensively used in commercial applications, where the efficiency of model serving is crucial for the user experience. Most current research focuses on optimizing individual sub-procedures, e.g. local inference and communication, however, there is no comprehensive framework that provides a holistic system view for optimizing LLM serving in an end-to-end manner. In this work, we conduct a detailed analysis to identify major bottlenecks that impact end-to-end latency in LLM serving systems. Our analysis reveals that a comprehensive LLM serving endpoint must address a series of efficiency bottlenecks that extend beyond LLM inference. We then propose ScaleLLM, an optimized system for resource-efficient LLM serving. Our extensive experiments reveal that with 64 concurrent requests, ScaleLLM achieves a 4.3x speed up over vLLM and outperforms state-of-the-arts with 1.5x higher throughput.

Large language models (LLMs) have demonstrated exceptional capabilities in generating text, images, and video content. However, as context length grows, the computational cost of attention increases quadratically with the number of tokens, presenting significant efficiency challenges. This paper presents an analysis of various Key-Value (KV) cache compression strategies, offering a comprehensive taxonomy that categorizes these methods by their underlying principles and implementation techniques. Furthermore, we evaluate their impact on performance and inference latency, providing critical insights into their effectiveness. Our findings highlight the trade-offs involved in KV cache compression and its influence on handling long-context scenarios, paving the way for more efficient LLM implementations.

With the ubiquitous use of modern large language models (LLMs) across industries, the inference serving for these models is ever expanding. Given the high compute and memory requirements of modern LLMs, more and more top-of-the-line GPUs are being deployed to serve these models. Energy availability has come to the forefront as the biggest challenge for data center expansion to serve these models. In this paper, we present the trade-offs brought up by making energy efficiency the primary goal of LLM serving under performance SLOs. We show that depending on the inputs, the model, and the service-level agreements, there are several knobs available to the LLM inference provider to use for being energy efficient. We characterize the impact of these knobs on the latency, throughput, as well as the energy. By exploring these trade-offs, we offer valuable insights into optimizing energy usage without compromising on performance, thereby paving the way for sustainable and cost-effective LLM deployment in data center environments.

Voice dictation is an increasingly important text input modality. Existing systems that allow both dictation and editing-by-voice restrict their command language to flat templates invoked by trigger words. In this work, we study the feasibility of allowing users to interrupt their dictation with spoken editing commands in open-ended natural language. We introduce a new task and dataset, TERTiUS, to experiment with such systems. To support this flexibility in real-time, a system must incrementally segment and classify spans of speech as either dictation or command, and interpret the spans that are commands. We experiment with using large pre-trained language models to predict the edited text, or alternatively, to predict a small text-editing program. Experiments show a natural trade-off between model accuracy and latency: a smaller model achieves 30% end-state accuracy with 1.3 seconds of latency, while a larger model achieves 55% end-state accuracy with 7 seconds of latency.

Efficiently serving neural network models with low latency is becoming more challenging due to increasing model complexity and parameter count. Model quantization offers a solution which simultaneously reduces memory footprint and compute requirements. However, aggressive quantization may lead to an unacceptable loss in model accuracy owing to differences in sensitivity to numerical imperfection across different layers in the model. To address this challenge, we propose a mixed-precision post training quantization (PTQ) approach that assigns different numerical precisions to tensors in a network based on their specific needs, for a reduced memory footprint and improved latency while preserving model accuracy. Previous works rely on layer-wise Hessian information to determine numerical precision, but as we demonstrate, Hessian estimation is typically insufficient in determining an effective ordering of layer sensitivities. We address this by augmenting the estimated Hessian with additional information to capture inter-layer dependencies. We demonstrate that this consistently improves PTQ performance along the accuracy-latency Pareto frontier across multiple models. Our method combines second-order information and inter-layer dependencies to guide a bisection search, finding quantization configurations within a user-configurable model accuracy degradation range. We evaluate the effectiveness of our method on the ResNet50, MobileNetV2, and BERT models. Our experiments demonstrate latency reductions compared to a 16-bit baseline of 25.48%, 21.69%, and 33.28% respectively, while maintaining model accuracy to within 99.99% of the baseline model.

We investigate an emerging setup in which a small, on-device language model (LM) with access to local data communicates with a frontier, cloud-hosted LM to solve real-world tasks involving financial, medical, and scientific reasoning over long documents. Can a local-remote collaboration reduce cloud inference costs while preserving quality? First, we consider a naive collaboration protocol where the local and remote models simply chat back and forth. Because only the local model reads the full context, this protocol achieves a 30.4x reduction in remote costs, but recovers only 87% of the performance of the frontier model. We identify two key limitations of this protocol: the local model struggles to (1) follow the remote model's multi-step instructions and (2) reason over long contexts. Motivated by these observations, we study an extension of this protocol, coined MinionS, in which the remote model decomposes the task into easier subtasks over shorter chunks of the document, that are executed locally in parallel. MinionS reduces costs by 5.7x on average while recovering 97.9% of the performance of the remote model alone. Our analysis reveals several key design choices that influence the trade-off between cost and performance in local-remote systems.

Structured width pruning of GLU-MLP layers, guided by the Maximum Absolute Weight (MAW) criterion, reveals a systematic dichotomy in how reducing the expansion ratio affects different model capabilities. While performance on tasks relying on parametric knowledge (e.g., MMLU, GSM8K) and perplexity metrics degrades predictably, instruction-following capabilities improve substantially (+46% to +75% in IFEval for Llama-3.2-1B and 3B models), and multi-step reasoning remains robust (MUSR). This pattern challenges the prevailing assumption that pruning induces uniform degradation. We evaluated seven expansion ratio configurations using comprehensive benchmarks assessing factual knowledge, mathematical reasoning, language comprehension, instruction-following, and truthfulness. Our analysis identifies the expansion ratio as a critical architectural parameter that selectively modulates cognitive capabilities, rather than merely serving as a compression metric. We provide the first systematic characterization of this selective preservation phenomenon. Notably, we document a robust inverse correlation (r = -0.864, p = 0.012 in Llama-3B) between factual knowledge capacity (MMLU) and truthfulness metrics (TruthfulQA-MC2): as knowledge degrades, the model's ability to discriminate misconceptions improves consistently. This connects two previously distinct research areas, demonstrating that MAW-guided width pruning acts as a selective filter, reducing parametric knowledge while preserving or enhancing behavioral alignment. Additionally, we quantify context-dependent efficiency trade-offs: pruned configurations achieve up to 23% reduction in energy consumption (J/token) but incur penalties in single-request latency, whereas batch processing workloads benefit uniformly.

Mixture of Experts (MoEs) have become a central component of many state-of-the-art open-source and proprietary large language models. Despite their widespread adoption, it remains unclear how close existing MoE architectures are to optimal with respect to inference cost, as measured by accuracy per floating-point operation and per parameter. In this work, we revisit MoE design from a hardware-software co-design perspective, grounded in empirical and theoretical considerations. We characterize key performance bottlenecks across diverse deployment regimes, spanning offline high-throughput execution and online, latency-critical inference. Guided by these insights, we introduce LatentMoE, a new model architecture resulting from systematic design exploration and optimized for maximal accuracy per unit of compute. Empirical design space exploration at scales of up to 95B parameters and over a 1T-token training horizon, together with supporting theoretical analysis, shows that LatentMoE consistently outperforms standard MoE architectures in terms of accuracy per FLOP and per parameter. Given its strong performance, the LatentMoE architecture has been adopted by the flagship Nemotron-3 Super and Ultra models and scaled to substantially larger regimes, including longer token horizons and larger model sizes, as reported in Nvidia et al. (arXiv:2512.20856).

Latency and efficiency issues are often overlooked when evaluating IR models based on Pretrained Language Models (PLMs) in reason of multiple hardware and software testing scenarios. Nevertheless, efficiency is an important part of such systems and should not be overlooked. In this paper, we focus on improving the efficiency of the SPLADE model since it has achieved state-of-the-art zero-shot performance and competitive results on TREC collections. SPLADE efficiency can be controlled via a regularization factor, but solely controlling this regularization has been shown to not be efficient enough. In order to reduce the latency gap between SPLADE and traditional retrieval systems, we propose several techniques including L1 regularization for queries, a separation of document/query encoders, a FLOPS-regularized middle-training, and the use of faster query encoders. Our benchmark demonstrates that we can drastically improve the efficiency of these models while increasing the performance metrics on in-domain data. To our knowledge, {we propose the first neural models that, under the same computing constraints, achieve similar latency (less than 4ms difference) as traditional BM25, while having similar performance (less than 10\% MRR@10 reduction) as the state-of-the-art single-stage neural rankers on in-domain data}.

DistServe improves the performance of large language models (LLMs) serving by disaggregating the prefill and decoding computation. Existing LLM serving systems colocate the two phases and batch the computation of prefill and decoding across all users and requests. We find that this strategy not only leads to strong prefill-decoding interferences but also couples the resource allocation and parallelism plans for both phases. LLM applications often emphasize individual latency for each phase: time to first token (TTFT) for the prefill phase and time per output token (TPOT) of each request for the decoding phase. In the presence of stringent latency requirements, existing systems have to prioritize one latency over the other, or over-provision compute resources to meet both. DistServe assigns prefill and decoding computation to different GPUs, hence eliminating prefill-decoding interferences. Given the application's TTFT and TPOT requirements, DistServe co-optimizes the resource allocation and parallelism strategy tailored for each phase. DistServe also places the two phases according to the serving cluster's bandwidth to minimize the communication caused by disaggregation. As a result, DistServe significantly improves LLM serving performance in terms of the maximum rate that can be served within both TTFT and TPOT constraints on each GPU. Our evaluations show that on various popular LLMs, applications, and latency requirements, DistServe can serve 4.48x more requests or 10.2x tighter SLO, compared to state-of-the-art systems, while staying within latency constraints for > 90% of requests.

We rethink test-time scaling laws from a practical efficiency perspective, revealing that the effectiveness of smaller models is significantly overestimated. Prior work, grounded in compute-optimality, overlooks critical memory access bottlenecks introduced by inference-time strategies (e.g., Best-of-N, long CoTs). Our holistic analysis, spanning models from 0.6B to 32B parameters, reveals a new Kinetics Scaling Law that better guides resource allocation by incorporating both computation and memory access costs. Kinetics Scaling Law suggests that test-time compute is more effective when used on models above a threshold than smaller ones. A key reason is that in TTS, attention, rather than parameter count, emerges as the dominant cost factor. Motivated by this, we propose a new scaling paradigm centered on sparse attention, which lowers per-token cost and enables longer generations and more parallel samples within the same resource budget. Empirically, we show that sparse attention models consistently outperform dense counterparts, achieving over 60 points gains in low-cost regimes and over 5 points gains in high-cost regimes for problem-solving accuracy on AIME, encompassing evaluations on state-of-the-art MoEs. These results suggest that sparse attention is essential for realizing the full potential of test-time scaling because, unlike training, where parameter scaling saturates, test-time accuracy continues to improve through increased generation. The code is available at https://github.com/Infini-AI-Lab/Kinetics.

Large Language Models (LLMs) have driven significant progress, yet their growing parameter counts and context windows incur prohibitive compute, energy, and monetary costs. We introduce EfficientLLM, a novel benchmark and the first comprehensive empirical study evaluating efficiency techniques for LLMs at scale. Conducted on a production-class cluster (48xGH200, 8xH200 GPUs), our study systematically explores three key axes: (1) architecture pretraining (efficient attention variants: MQA, GQA, MLA, NSA; sparse Mixture-of-Experts (MoE)), (2) fine-tuning (parameter-efficient methods: LoRA, RSLoRA, DoRA), and (3) inference (quantization methods: int4, float16). We define six fine-grained metrics (Memory Utilization, Compute Utilization, Latency, Throughput, Energy Consumption, Compression Rate) to capture hardware saturation, latency-throughput balance, and carbon cost. Evaluating over 100 model-technique pairs (0.5B-72B parameters), we derive three core insights: (i) Efficiency involves quantifiable trade-offs: no single method is universally optimal; e.g., MoE reduces FLOPs and improves accuracy but increases VRAM by 40%, while int4 quantization cuts memory/energy by up to 3.9x at a 3-5% accuracy drop. (ii) Optima are task- and scale-dependent: MQA offers optimal memory-latency trade-offs for constrained devices, MLA achieves lowest perplexity for quality-critical tasks, and RSLoRA surpasses LoRA efficiency only beyond 14B parameters. (iii) Techniques generalize across modalities: we extend evaluations to Large Vision Models (Stable Diffusion 3.5, Wan 2.1) and Vision-Language Models (Qwen2.5-VL), confirming effective transferability. By open-sourcing datasets, evaluation pipelines, and leaderboards, EfficientLLM provides essential guidance for researchers and engineers navigating the efficiency-performance landscape of next-generation foundation models.

Despite their remarkable success in complex tasks propelling widespread adoption, large language-model-based agents still face critical deployment challenges due to prohibitive latency and inference costs. While recent work has explored various methods to accelerate inference, existing approaches suffer from significant limitations: they either fail to preserve performance fidelity, require extensive offline training of router modules, or incur excessive operational costs. Moreover, they provide minimal user control over the tradeoff between acceleration and other performance metrics. To address these gaps, we introduce Dynamic Speculative Planning (DSP), an asynchronous online reinforcement learning framework that provides lossless acceleration with substantially reduced costs without requiring additional pre-deployment preparation. DSP explicitly optimizes a joint objective balancing end-to-end latency against dollar cost, allowing practitioners to adjust a single parameter that steers the system toward faster responses, cheaper operation, or any point along this continuum. Experiments on two standard agent benchmarks demonstrate that DSP achieves comparable efficiency to the fastest lossless acceleration method while reducing total cost by 30% and unnecessary cost up to 60%. Our code and data are available through https://github.com/guanyilin428/Dynamic-Speculative-Planning.

Auto-regressive Large Language Models (LLMs) demonstrate remarkable performance across different domains such as vision and language processing. However, due to sequential processing through a stack of transformer layers, autoregressive decoding faces significant computation/latency challenges, particularly in resource-constrained environments like mobile and edge devices. Existing approaches in literature that aim to improve latency via skipping layers have two distinct flavors - 1) Early exit, and 2) Input-agnostic heuristics where tokens exit at pre-determined layers irrespective of input sequence. Both the above strategies have limitations - the former cannot be applied to handle KV Caching necessary for speed-ups in modern framework and the latter does not capture the variation in layer importance across tasks or more generally, across input sequences. To address both limitations, we propose FiRST, an algorithm that reduces inference latency by using layer-specific routers to select a subset of transformer layers adaptively for each input sequence - the prompt (during the prefill stage) decides which layers will be skipped during decoding. FiRST preserves compatibility with KV caching enabling faster inference while being quality-aware. FiRST is model-agnostic and can be easily enabled on any pre-trained LLM. Our approach reveals that input adaptivity is critical - indeed, different task-specific middle layers play a crucial role in evolving hidden representations depending on tasks. Extensive experiments show that FiRST significantly reduces latency while outperforming other layer selection strategies in quality metics. It retains competitive performance to base model (without layer skipping) and in some cases, even improves upon it. FiRST is thus a promising and efficient solution for LLM deployment in low-resource environments.

State-of-the-art deep learning models have achieved significant performance levels on various benchmarks. However, the excellent performance comes at a cost of inefficient computational cost. Light-weight architectures, on the other hand, achieve moderate accuracies, but at a much more desirable latency. This paper presents a new method of jointly using the large accurate models together with the small fast ones. To this end, we propose an Energy-Based Joint Reasoning (EBJR) framework that adaptively distributes the samples between shallow and deep models to achieve an accuracy close to the deep model, but latency close to the shallow one. Our method is applicable to out-of-the-box pre-trained models as it does not require an architecture change nor re-training. Moreover, it is easy to use and deploy, especially for cloud services. Through a comprehensive set of experiments on different down-stream tasks, we show that our method outperforms strong state-of-the-art approaches with a considerable margin. In addition, we propose specialized EBJR, an extension of our method where we create a smaller specialized side model that performs the target task only partially, but yields an even higher accuracy and faster inference. We verify the strengths of our methods with both theoretical and experimental evaluations.

Scaling language models to long contexts is often bottlenecked by the size of the key-value (KV) cache. In deployed settings, long contexts are typically managed through compaction in token space via summarization. However, summarization can be highly lossy, substantially harming downstream performance. Recent work on Cartridges has shown that it is possible to train highly compact KV caches in latent space that closely match full-context performance, but at the cost of slow and expensive end-to-end optimization. This work describes an approach for fast context compaction in latent space through Attention Matching, which constructs compact keys and values to reproduce attention outputs and preserve attention mass at a per-KV-head level. We show that this formulation naturally decomposes into simple subproblems, some of which admit efficient closed-form solutions. Within this framework, we develop a family of methods that significantly push the Pareto frontier of compaction time versus quality, achieving up to 50x compaction in seconds on some datasets with little quality loss.

In simultaneous speech translation (SimulST), finding the best trade-off between high translation quality and low latency is a challenging task. To meet the latency constraints posed by the different application scenarios, multiple dedicated SimulST models are usually trained and maintained, generating high computational costs. In this paper, motivated by the increased social and environmental impact caused by these costs, we investigate whether a single model trained offline can serve not only the offline but also the simultaneous task without the need for any additional training or adaptation. Experiments on en->{de, es} indicate that, aside from facilitating the adoption of well-established offline techniques and architectures without affecting latency, the offline solution achieves similar or better translation quality compared to the same model trained in simultaneous settings, as well as being competitive with the SimulST state of the art.

Models for low-latency, streaming applications could benefit from the knowledge capacity of larger models, but edge devices cannot run these models due to resource constraints. A possible solution is to transfer hints during inference from a large model running remotely to a small model running on-device. However, this incurs a communication delay that breaks real-time requirements and does not guarantee that both models will operate on the same data at the same time. We propose knowledge boosting, a novel technique that allows a large model to operate on time-delayed input during inference, while still boosting small model performance. Using a streaming neural network that processes 8 ms chunks, we evaluate different speech separation and enhancement tasks with communication delays of up to six chunks or 48 ms. Our results show larger gains where the performance gap between the small and large models is wide, demonstrating a promising method for large-small model collaboration for low-latency applications. Code, dataset, and audio samples available at https://knowledgeboosting.cs.washington.edu/.

Modern semantic search and retrieval-augmented generation (RAG) systems rely predominantly on in-memory approximate nearest neighbor (ANN) indexes over high-precision floating-point vectors, resulting in escalating operational cost and inherent trade-offs between latency, throughput, and retrieval accuracy. This paper analyzes the architectural limitations of the dominant "HNSW + float32 + cosine similarity" stack and evaluates existing cost-reduction strategies, including storage disaggregation and lossy vector quantization, which inevitably sacrifice either performance or accuracy. We introduce and empirically evaluate an alternative information-theoretic architecture based on maximally informative binarization (MIB), efficient bitwise distance metrics, and an information-theoretic scoring (ITS) mechanism. Unlike conventional ANN systems, this approach enables exhaustive search over compact binary representations, allowing deterministic retrieval and eliminating accuracy degradation under high query concurrency. Using the MAIR benchmark across 14 datasets and 10,038 queries, we compare this architecture against Elasticsearch, Pinecone, PGVector, and Qdrant. Results demonstrate retrieval quality comparable to full-precision systems, while achieving substantially lower latency and maintaining constant throughput at high request rates. We show that this architectural shift enables a truly serverless, cost-per-query deployment model, challenging the necessity of large in-memory ANN indexes for high-quality semantic search.

AI Agents can perform complex operations at great speed, but just like all the humans we have ever hired, their intelligence remains fallible. Miscommunications aren't noticed, systemic biases have no counter-action, and inner monologues are rarely written down. We did not come to fire them for their mistakes, but to hire them and provide a safe productive working environment. We posit that we can reuse a common corporate organizational structure: teams of independent AI agents with strict role boundaries can work with common goals, but opposing incentives. Multiple models serving as a team of rivals can catch and minimize errors within the final product at a small cost to the velocity of actions. In this paper we demonstrate that we can achieve reliability without acquiring perfect components, but through careful orchestration of imperfect ones. This paper describes the architecture of such a system in practice: specialized agent teams (planners, executors, critics, experts), organized into an organization with clear goals, coordinated through a remote code executor that keeps data transformations and tool invocations separate from reasoning models. Rather than agents directly calling tools and ingesting full responses, they write code that executes remotely; only relevant summaries return to agent context. By preventing raw data and tool outputs from contaminating context windows, the system maintains clean separation between perception (brains that plan and reason) and execution (hands that perform heavy data transformations and API calls). We demonstrate the approach achieves over 90% internal error interception prior to user exposure while maintaining acceptable latency tradeoffs. A survey from our traces shows that we only trade off cost and latency to achieve correctness and incrementally expand capabilities without impacting existing ones.

Recent advances in retrieval models based on learned sparse representations generated by transformers have led us to, once again, consider score-at-a-time query evaluation techniques for the top-k retrieval problem. Previous studies comparing document-at-a-time and score-at-a-time approaches have consistently found that the former approach yields lower mean query latency, although the latter approach has more predictable query latency. In our experiments with four different retrieval models that exploit representational learning with bags of words, we find that transformers generate "wacky weights" that appear to greatly reduce the opportunities for skipping and early exiting optimizations that lie at the core of standard document-at-a-time techniques. As a result, score-at-a-time approaches appear to be more competitive in terms of query evaluation latency than in previous studies. We find that, if an effectiveness loss of up to three percent can be tolerated, a score-at-a-time approach can yield substantial gains in mean query latency while at the same time dramatically reducing tail latency.

Large Language Models (LLMs) demonstrate substantial accuracy gains when augmented with reasoning modes such as chain-of-thought and inference-time scaling. However, reasoning also incurs significant costs in inference latency and token usage, with environmental and financial impacts, which are unnecessary for many simple prompts. We present a semantic router that classifies queries based on their reasoning requirements and selectively applies reasoning only when beneficial. Our approach achieves a 10.2 percentage point improvement in accuracy on the MMLU-Pro benchmark while reducing response latency by 47.1% and token consumption by 48.5% compared to direct inference with vLLM. These results demonstrate that semantic routing offers an effective mechanism for striking a balance between accuracy and efficiency in open-source LLM serving systems

One common strategy for improving the performance of Large Language Models (LLMs) on downstream tasks involves using a verifier model to either select the best answer from a pool of candidates or to steer the auto-regressive generation process towards better outputs. This class of methods typically results in improved accuracy at the cost of increased computation at test-time, a paradigm known as test-time scaling. However, most existing approaches increase computation uniformly across all samples and generation steps, without considering the complexity of individual instances, leading to inefficient resource use. We address this limitation by proposing an approach, called Locally Adaptive Test-Time Scaling (LATTS), that allocates variable compute across generation steps. Specifically, at each generation step, LATTS employs a verifier-based acceptance criterion to decide whether to resample, backtrack, restart, or stop the generation process. This criterion effectively adjusts the per-step computational effort based on a precise notion of local difficulty derived from the verifier model. Empirical results show that LATTS achieves significantly superior accuracy--compute tradeoffs compared to standard verifier-based methods.

As large language models (LLMs) increasingly tackle complex reasoning tasks, test-time scaling has become critical for enhancing capabilities. However, in agentic scenarios with frequent tool calls, the traditional generation-length-based definition breaks down: tool latency decouples inference time from generation length. We propose Timely Machine, redefining test-time as wall-clock time, where models dynamically adjust strategies based on time budgets. We introduce Timely-Eval, a benchmark spanning high-frequency tool calls, low-frequency tool calls, and time-constrained reasoning. By varying tool latency, we find smaller models excel with fast feedback through more interactions, while larger models dominate high-latency settings via superior interaction quality. Moreover, existing models fail to adapt reasoning to time budgets. We propose Timely-RL to address this gap. After cold-start supervised fine-tuning, we use reinforcement learning to enhance temporal planning. Timely-RL improves time budget awareness and consistently boosts performance across Timely-Eval. We hope our work offers a new perspective on test-time scaling for the agentic era.

Neural architecture search automates neural network design and has achieved state-of-the-art results in many deep learning applications. While recent literature has focused on designing networks to maximize accuracy, little work has been conducted to understand the compatibility of architecture design spaces to varying hardware. In this paper, we analyze the neural blocks used to build Once-for-All (MobileNetV3), ProxylessNAS and ResNet families, in order to understand their predictive power and inference latency on various devices, including Huawei Kirin 9000 NPU, RTX 2080 Ti, AMD Threadripper 2990WX, and Samsung Note10. We introduce a methodology to quantify the friendliness of neural blocks to hardware and the impact of their placement in a macro network on overall network performance via only end-to-end measurements. Based on extensive profiling results, we derive design insights and apply them to hardware-specific search space reduction. We show that searching in the reduced search space generates better accuracy-latency Pareto frontiers than searching in the original search spaces, customizing architecture search according to the hardware. Moreover, insights derived from measurements lead to notably higher ImageNet top-1 scores on all search spaces investigated.

Large language models (LLMs) have excelled in various applications, yet serving them at scale is challenging due to their substantial resource demands and high latency. Our real-world studies reveal that over 70% of user requests to LLMs have semantically similar counterparts, suggesting the potential for knowledge transfer among requests. However, naively caching and reusing past responses leads to a big quality drop. In this paper, we introduce IC-Cache, a caching system that enables live LLM capability augmentation to improve serving efficiency: by leveraging historical request-response pairs from larger models as in-context examples, IC-Cache empowers small LLMs to imitate and even exceed the compositional abilities (e.g., reasoning) of their larger counterparts, enabling selective offloading of requests to reduce cost and latency. Achieving this live augmentation at scale introduces intricate trade-offs between response quality, latency, and system throughput. For a new request, IC-Cache efficiently selects similar, high-utility examples to prepend them to the new request's input. At scale, it adaptively routes requests across LLMs of varying capabilities, accounting for response quality and serving loads. IC-Cache employs a cost-aware cache replay mechanism that refines example quality offline to maximize online cache utility and efficiency. Evaluations on millions of realistic requests demonstrate that IC-Cache improves LLM serving throughput by 1.4-5.9x and reduces latency by 28-71% without hurting response quality.

Simultaneous translation of unbounded streaming speech remains a challenging problem due to the need for effectively processing the history speech context and past translations so that quality and latency, including computation overhead, can be balanced. Most prior works assume pre-segmented speech, limiting their real-world applicability. In this paper, we propose InfiniSST, a novel approach that formulates SST as a multi-turn dialogue task, enabling seamless translation of unbounded speech. We construct translation trajectories and robust segments from MuST-C with multi-latency augmentation during training and develop a key-value (KV) cache management strategy to facilitate efficient inference. Experiments on MuST-C En-Es, En-De, and En-Zh demonstrate that InfiniSST reduces computation-aware latency by 0.5 to 1 second while maintaining the same translation quality compared to baselines. Ablation studies further validate the contributions of our data construction and cache management strategy. We release the code and demo at https://github.com/LeiLiLab/InfiniSST

Production-grade LLM systems require robust adherence to dozens or even hundreds of instructions simultaneously. However, the instruction-following capabilities of LLMs at high instruction densities have not yet been characterized, as existing benchmarks only evaluate models on tasks with a single or few instructions. We introduce IFScale, a simple benchmark of 500 keyword-inclusion instructions for a business report writing task to measure how instruction-following performance degrades as instruction density increases. We evaluate 20 state-of-the-art models across seven major providers and find that even the best frontier models only achieve 68% accuracy at the max density of 500 instructions. Our analysis reveals model size and reasoning capability to correlate with 3 distinct performance degradation patterns, bias towards earlier instructions, and distinct categories of instruction-following errors. Our insights can help inform design of instruction-dense prompts in real-world applications and highlight important performance-latency tradeoffs. We open-source the benchmark and all results for further analysis at https://distylai.github.io/IFScale.

Large language model (LLM) queries are predominantly processed by frontier models in centralized cloud infrastructure. Rapidly growing demand strains this paradigm, and cloud providers struggle to scale infrastructure at pace. Two advances enable us to rethink this paradigm: small LMs (<=20B active parameters) now achieve competitive performance to frontier models on many tasks, and local accelerators (e.g., Apple M4 Max) run these models at interactive latencies. This raises the question: can local inference viably redistribute demand from centralized infrastructure? Answering this requires measuring whether local LMs can accurately answer real-world queries and whether they can do so efficiently enough to be practical on power-constrained devices (i.e., laptops). We propose intelligence per watt (IPW), task accuracy divided by unit of power, as a metric for assessing capability and efficiency of local inference across model-accelerator pairs. We conduct a large-scale empirical study across 20+ state-of-the-art local LMs, 8 accelerators, and a representative subset of LLM traffic: 1M real-world single-turn chat and reasoning queries. For each query, we measure accuracy, energy, latency, and power. Our analysis reveals 3 findings. First, local LMs can accurately answer 88.7% of single-turn chat and reasoning queries with accuracy varying by domain. Second, from 2023-2025, IPW improved 5.3x and local query coverage rose from 23.2% to 71.3%. Third, local accelerators achieve at least 1.4x lower IPW than cloud accelerators running identical models, revealing significant headroom for optimization. These findings demonstrate that local inference can meaningfully redistribute demand from centralized infrastructure, with IPW serving as the critical metric for tracking this transition. We release our IPW profiling harness for systematic intelligence-per-watt benchmarking.

Latency-critical speech applications (e.g., live transcription, voice commands, and real-time translation) demand low time-to-first-token (TTFT) and high transcription accuracy, particularly on resource-constrained edge devices. Full-attention Transformer encoders remain a strong accuracy baseline for automatic speech recognition (ASR) because every frame can directly attend to every other frame, which resolves otherwise locally ambiguous acoustics using distant lexical context. However, this global dependency incurs quadratic complexity in sequence length, inducing an inherent "encode-the-whole-utterance" latency profile. For streaming use cases, this causes TTFT to grow linearly with utterance length as the encoder must process the entire prefix before any decoder token can be emitted. To better meet the needs of on-device, streaming ASR use cases we introduce Moonshine v2, an ergodic streaming-encoder ASR model that employs sliding-window self-attention to achieve bounded, low-latency inference while preserving strong local context. Our models achieve state of the art word error rates across standard benchmarks, attaining accuracy on-par with models 6x their size while running significantly faster. These results demonstrate that carefully designed local attention is competitive with the accuracy of full attention at a fraction of the size and latency cost, opening new possibilities for interactive speech interfaces on edge devices.

Vector Similarity Search (VSS) in high-dimensional spaces is rapidly emerging as core functionality in next-generation database systems for numerous data-intensive services -- from embedding lookups in large language models (LLMs), to semantic information retrieval and recommendation engines. Current benchmarks, however, evaluate VSS primarily on the recall-latency trade-off against a ground truth defined solely by distance metrics, neglecting how retrieval quality ultimately impacts downstream tasks. This disconnect can mislead both academic research and industrial practice. We present Iceberg, a holistic benchmark suite for end-to-end evaluation of VSS methods in realistic application contexts. From a task-centric view, Iceberg uncovers the Information Loss Funnel, which identifies three principal sources of end-to-end performance degradation: (1) Embedding Loss during feature extraction; (2) Metric Misuse, where distances poorly reflect task relevance; (3) Data Distribution Sensitivity, highlighting index robustness across skews and modalities. For a more comprehensive assessment, Iceberg spans eight diverse datasets across key domains such as image classification, face recognition, text retrieval, and recommendation systems. Each dataset, ranging from 1M to 100M vectors, includes rich, task-specific labels and evaluation metrics, enabling assessment of retrieval algorithms within the full application pipeline rather than in isolation. Iceberg benchmarks 13 state-of-the-art VSS methods and re-ranks them based on application-level metrics, revealing substantial deviations from traditional rankings derived purely from recall-latency evaluations. Building on these insights, we define a set of task-centric meta-features and derive an interpretable decision tree to guide practitioners in selecting and tuning VSS methods for their specific workloads.

Instruction finetuning is standard practice for improving LLM performance, yet it remains unclear whether it enhances reasoning or merely induces surface-level pattern matching. We investigate this by evaluating base and instruction-tuned models on standard math benchmarks, structurally perturbed variants, and domain-shifted tasks. Our analysis highlights two key (often overlooked) limitations of instruction tuning. First, the performance advantage is unstable and depends heavily on evaluation settings. In zero-shot CoT settings on GSM8K, base models consistently outperform instruction-tuned variants, with drops as high as 32.67\% (Llama3-70B). Instruction-tuned models only match or exceed this performance when provided with few-shot exemplars, suggesting a reliance on specific prompting patterns rather than intrinsic reasoning. Second, tuning gains are brittle under distribution shift. Our results show that base models surpass instruction-tuned variants on the domain-specific MedCalc benchmark. Additionally, instruction-tuned models show sharp declines on perturbed datasets, indicating sensitivity to prompt structure over robust reasoning.

Large reasoning models (LRMs) have exhibited the capacity of enhancing reasoning performance via internal test-time scaling. Building upon this, a promising direction is to further scale test-time compute to unlock even greater reasoning capabilities. However, as we push these scaling boundaries, systematically understanding the practical limits and achieving optimal resource allocation becomes a critical challenge. In this paper, we investigate the scaling Pareto of test-time scaling and introduce the Test-Time Scaling Performance Model (TTSPM). We theoretically analyze two fundamental paradigms for such extended scaling, parallel scaling and sequential scaling, from a probabilistic modeling perspective. Our primary contribution is the derivation of the saturation point on the scaling budget for both strategies, identifying thresholds beyond which additional computation yields diminishing returns. Remarkably, despite their distinct mechanisms, both paradigms converge to a unified mathematical structure in their upper bounds. We empirically validate our theoretical findings on challenging reasoning benchmarks, including AIME, MATH-500, and GPQA, demonstrating the practical utility of these bounds for test-time resource allocation. We hope that this work provides insights into the cost-benefit trade-offs of test-time scaling, guiding the development of more resource-efficient inference strategies for large reasoning models.

While 4-bit quantization has emerged as a memory-optimal choice for non-reasoning models and zero-shot tasks across scales, we show that this universal prescription fails for reasoning models, where the KV cache rather than model size can dominate memory. Through systematic experiments across 1,700 inference scenarios on AIME25 and GPQA-Diamond, we find a scale-dependent trade-off: models with an effective size below 8-bit 4B parameters achieve better accuracy by allocating memory to more weights rather than longer generation, while larger models achieve better accuracy by allocating memory to longer generations. This scale threshold also determines when parallel scaling becomes memory-efficient and whether KV cache eviction outperforms KV quantization. Our findings show that memory optimization for LLMs cannot be scale-agnostic, while providing principled guidelines: for small reasoning models, prioritize model capacity over test-time compute, while for larger ones, maximize test-time compute. Our results suggest that optimizing reasoning models for deployment requires fundamentally different strategies from those established for non-reasoning models.

One technique to improve the retrieval effectiveness of a search engine is to expand documents with terms that are related or representative of the documents' content.From the perspective of a question answering system, this might comprise questions the document can potentially answer. Following this observation, we propose a simple method that predicts which queries will be issued for a given document and then expands it with those predictions with a vanilla sequence-to-sequence model, trained using datasets consisting of pairs of query and relevant documents. By combining our method with a highly-effective re-ranking component, we achieve the state of the art in two retrieval tasks. In a latency-critical regime, retrieval results alone (without re-ranking) approach the effectiveness of more computationally expensive neural re-rankers but are much faster.

Large Language Models (LLMs), constrained by their auto-regressive nature, suffer from slow decoding. Speculative decoding methods have emerged as a promising solution to accelerate LLM decoding, attracting attention from both systems and AI research communities. Recently, the pursuit of better draft quality has driven a trend toward parametrically larger draft models, which inevitably introduces substantial computational overhead. While existing work attempts to balance the trade-off between prediction accuracy and compute latency, we address this fundamental dilemma through architectural innovation. We propose PRISM, which disaggregates the computation of each predictive step across different parameter sets, refactoring the computational pathways of draft models to successfully decouple model capacity from inference cost. Through extensive experiments, we demonstrate that PRISM outperforms all existing draft architectures, achieving exceptional acceptance lengths while maintaining minimal draft latency for superior end-to-end speedup. We also re-examine scaling laws with PRISM, revealing that PRISM scales more effectively with expanding data volumes than other draft architectures. Through rigorous and fair comparison, we show that PRISM boosts the decoding throughput of an already highly optimized inference engine by more than 2.6x.

State-of-the-art neural network architectures such as ResNet, MobileNet, and DenseNet have achieved outstanding accuracy over low MACs and small model size counterparts. However, these metrics might not be accurate for predicting the inference time. We suggest that memory traffic for accessing intermediate feature maps can be a factor dominating the inference latency, especially in such tasks as real-time object detection and semantic segmentation of high-resolution video. We propose a Harmonic Densely Connected Network to achieve high efficiency in terms of both low MACs and memory traffic. The new network achieves 35%, 36%, 30%, 32%, and 45% inference time reduction compared with FC-DenseNet-103, DenseNet-264, ResNet-50, ResNet-152, and SSD-VGG, respectively. We use tools including Nvidia profiler and ARM Scale-Sim to measure the memory traffic and verify that the inference latency is indeed proportional to the memory traffic consumption and the proposed network consumes low memory traffic. We conclude that one should take memory traffic into consideration when designing neural network architectures for high-resolution applications at the edge.

Large Language Models (LLMs) exhibit potential artificial generic intelligence recently, however, their usage is costly with high response latency. Given mixed LLMs with their own strengths and weaknesses, LLM routing aims to identify the most suitable model for each query in the stream to maximize response quality and minimize cost and latency. However, the challenges involve: (1) dynamic trade-offs among quality, cost, and latency; (2) enabling continual learning in deployed systems; and (3) navigating a varying (e.g., new LLM addition or old LLM removal) set of LLM candidates over time. To bridge these gaps, we develop MixLLM, a dynamic contextual-bandit-based routing system for query-LLM assignment. Specifically, we first leverage query tags to enhance query embeddings for the routing task. Next, we design lightweight prediction models to estimate the response qualities and costs of queries over LLMs. We then devise a meta-decision maker to choose the query-LLM assignments to best tradeoff response quality, cost, and latency. Finally, the system benefits from continual training, allowing it to adapt to evolving queries and user feedback over time. Our extensive experiments show that MixLLM achieves the best trade-offs in response quality, cost, and latency (97.25% of GPT-4's quality at 24.18% of the cost under the time constraint).

As large language models (LLMs) take on more complex tasks, their inputs incorporate longer contexts to respond to questions that require domain knowledge or user-specific conversational histories. Yet, using long contexts poses a challenge for responsive LLM systems, as nothing can be generated until all the contexts are fetched to and processed by the LLM. Existing systems optimize only the computation delay in context processing (e.g., by caching intermediate key-value features of the text context) but often cause longer network delays in context fetching (e.g., key-value features consume orders of magnitude larger bandwidth than the text context). This paper presents CacheGen to minimize the delays in fetching and processing contexts for LLMs. CacheGen reduces the bandwidth needed for transmitting long contexts' key-value (KV) features through a novel encoder that compresses KV features into more compact bitstream representations. The encoder combines adaptive quantization with a tailored arithmetic coder, taking advantage of the KV features' distributional properties, such as locality across tokens. Furthermore, CacheGen minimizes the total delay in fetching and processing a context by using a controller that determines when to load the context as compressed KV features or raw text and picks the appropriate compression level if loaded as KV features. We test CacheGen on three models of various sizes and three datasets of different context lengths. Compared to recent methods that handle long contexts, CacheGen reduces bandwidth usage by 3.7-4.3x and the total delay in fetching and processing contexts by 2.7-3x while maintaining similar LLM performance on various tasks as loading the text contexts.

We introduce a novel framework for analyzing sorting algorithms in pairwise ranking prompting (PRP), re-centering the cost model around LLM inferences rather than traditional pairwise comparisons. While classical metrics based on comparison counts have traditionally been used to gauge efficiency, our analysis reveals that expensive LLM inferences overturn these predictions; accordingly, our framework encourages strategies such as batching and caching to mitigate inference costs. We show that algorithms optimal in the classical setting can lose efficiency when LLM inferences dominate the cost under certain optimizations.

Parallel thinking enhances LLM reasoning by multi-path sampling and aggregation. In system-level evaluations, a global parallelism level N is allocated to all samples, typically set large to maximize overall dataset accuracy. However, due to sample heterogeneity, some samples can achieve comparable performance with a smaller N'< N, causing budget redundancy. This incompatibility between system-level efficacy and sample-level efficiency constitutes the overscaling curse. In this paper, we formalize and quantify the overscaling curse, showing its universality and severity in practice, and analyze its trigger mechanism. We then propose a lightweight method, T2, to break the overscaling curse, which utilizes latent representations to estimate the optimal parallelism level for each sample before decoding. Experiments show that T2 significantly reduces cost while maintaining comparable performance, enabling more efficient parallel thinking.

Simultaneous Speech Translation (SimulST) systems stream in audio while simultaneously emitting translated text or speech. Such systems face the significant challenge of balancing translation quality and latency. We introduce a strategy to optimize this tradeoff: wait for more input only if you gain information by doing so. Based on this strategy, we present Regularized Entropy INformation Adaptation (REINA), a novel loss to train an adaptive policy using an existing non-streaming translation model. We derive REINA from information theory principles and show that REINA helps push the reported Pareto frontier of the latency/quality tradeoff over prior works. Utilizing REINA, we train a SimulST model on French, Spanish and German, both from and into English. Training on only open source or synthetically generated data, we achieve state-of-the-art (SOTA) streaming results for models of comparable size. We also introduce a metric for streaming efficiency, quantitatively showing REINA improves the latency/quality trade-off by as much as 21% compared to prior approaches, normalized against non-streaming baseline BLEU scores.

The standard paradigm for solving coding tasks via large language models (LLMs) is to generate-then-rank programs, where the latter step uses a verifier in the ranking process. The growing consensus is that a comprehensive verifier (e.g., a full test suite) should be prioritized over an outcome reward model (ORM) whenever possible, with little consideration given to the trade-offs involved. We aim to challenge this assumption by systematically exploring the tradeoff between speed and accuracy. We find that ORMs play a crucial role in scaling verification through trading accuracy for speed, even when a comprehensive verifier is available. Their value becomes especially apparent when used in a generate-prune-then-rank approach, where a faster but less accurate verifier removes incorrect solutions prior to ranking -- leading to a system that is 11.65x faster while only being 8.33% less accurate than the full test suite. We analyze the generate-prune-then-rank approach and show that it works by filtering out incorrect but highly ranked solutions. These findings enable the design of scalable and accurate program ranking systems.

AI Video Assistant emerges as a new paradigm for Real-time Communication (RTC), where one peer is a Multimodal Large Language Model (MLLM) deployed in the cloud. This makes interaction between humans and AI more intuitive, akin to chatting with a real person. However, a fundamental mismatch exists between current RTC frameworks and AI Video Assistants, stemming from the drastic shift in Quality of Experience (QoE) and more challenging networks. Measurements on our production prototype also confirm that current RTC fails, causing latency spikes and accuracy drops. To address these challenges, we propose Artic, an AI-oriented RTC framework for MLLM Video Assistants, exploring the shift from "humans watching video" to "AI understanding video." Specifically, Artic proposes: (1) Response Capability-aware Adaptive Bitrate, which utilizes MLLM accuracy saturation to proactively cap bitrate, reserving bandwidth headroom to absorb future fluctuations for latency reduction; (2) Zero-overhead Context-aware Streaming, which allocates limited bitrate to regions most important for the response, maintaining accuracy even under ultra-low bitrates; and (3) Degraded Video Understanding Benchmark, the first benchmark evaluating how RTC-induced video degradation affects MLLM accuracy. Prototype experiments using real-world uplink traces show that compared with existing methods, Artic significantly improves accuracy by 15.12% and reduces latency by 135.31 ms. We will release the benchmark and codes at https://github.com/pku-netvideo/DeViBench.

Since the breakthrough performance of AlexNet in 2012, convolutional neural networks (convnets) have grown into extremely powerful vision models. Deep learning researchers have used convnets to perform vision tasks with accuracy that was unachievable a decade ago. Confronted with the immense computation that convnets use, deep learning researchers also became interested in efficiency. However, the engineers who deployed efficient convnets soon realized that they were slower than the previous generation, despite using fewer operations. Many reverted to older models that ran faster. Hence researchers switched the objective of their search from arithmetic complexity to latency and produced a new wave of models that performed better. Paradoxically, these models also used more operations. Skepticism grew among researchers and engineers alike about the relevance of arithmetic complexity. Contrary to the prevailing view that latency and arithmetic complexity are irreconcilable, a simple formula relates both through computational efficiency. This insight enabled us to co-optimize the separate factors that determine latency. We observed that the degenerate conv2d layers that produce the best accuracy--complexity trade-off also use significant memory resources and have low computational efficiency. We devised block fusion algorithms to implement all the layers of a residual block in a single kernel, thereby creating temporal locality, avoiding communication, and reducing workspace size. Our ConvFirst model with block-fusion kernels has less arithmetic complexity and greater computational efficiency than baseline models and kernels, and ran approximately four times as fast as ConvNeXt. We also created novel tools, including efficiency gap plots and waterline analysis. Our unified approach to convnet efficiency envisions a new era of models and kernels that achieve greater accuracy at lower cost.

The generation speed of LLMs are bottlenecked by autoregressive decoding, where tokens are predicted sequentially one by one. Alternatively, diffusion large language models (dLLMs) theoretically allow for parallel token generation, but in practice struggle to achieve the speed of autoregressive models without significantly sacrificing quality. We therefore introduce adaptive parallel decoding (APD), a novel method that dynamically adjusts the number of tokens sampled in parallel. We achieve this by defining a multiplicative mixture between the dLLM marginal probabilities and the joint probability of sequences under a small auxiliary autoregressive model. This inverts the standard setup of speculative decoding, where the goal is to sample from a large autoregressive verifier by drafting from a smaller model. We further optimize APD by enabling KV caching and limiting the size of the masked input. Altogether, our method puts forward three tunable parameters to flexibly tradeoff throughput and quality. We show that APD provides markedly higher throughput with minimal quality degradations on downstream benchmarks.

Large language models (LLMs) exhibit impressive capabilities across a wide range of tasks, yet the choice of which model to use often involves a trade-off between performance and cost. More powerful models, though effective, come with higher expenses, while less capable models are more cost-effective. To address this dilemma, we propose several efficient router models that dynamically select between a stronger and a weaker LLM during inference, aiming to optimize the balance between cost and response quality. We develop a training framework for these routers leveraging human preference data and data augmentation techniques to enhance performance. Our evaluation on widely-recognized benchmarks shows that our approach significantly reduces costs-by over 2 times in certain cases-without compromising the quality of responses. Interestingly, our router models also demonstrate significant transfer learning capabilities, maintaining their performance even when the strong and weak models are changed at test time. This highlights the potential of these routers to provide a cost-effective yet high-performance solution for deploying LLMs.

Inference with transformer-based language models begins with a prompt processing step. In this step, the model generates the first output token and stores the KV cache needed for future generation steps. This prompt processing step can be computationally expensive, taking 10s of seconds or more for billion-parameter models on edge devices when prompt lengths or batch sizes rise. This degrades user experience by introducing significant latency into the model's outputs. To reduce the time spent producing the first output (known as the ``time to first token'', or TTFT) of a pretrained model, we introduce a novel method called KV Prediction. In our method, a small auxiliary model is used to process the prompt and produce an approximation of the KV cache used by a base model. This approximated KV cache is then used with the base model for autoregressive generation without the need to query the auxiliary model again. We demonstrate that our method produces a pareto-optimal efficiency-accuracy trade-off when compared to baselines. On TriviaQA, we demonstrate relative accuracy improvements in the range of 15%-50% across a range of TTFT FLOPs budgets. We also demonstrate accuracy improvements of up to 30% on HumanEval python code completion at fixed TTFT FLOPs budgets. Additionally, we benchmark models on an Apple M2 Pro CPU and demonstrate that our improvement in FLOPs translates to a TTFT speedup on hardware. We release our code at https://github.com/apple/corenet/tree/main/projects/kv-prediction .

Parallel Speculative Decoding (PSD) accelerates traditional Speculative Decoding (SD) by overlapping draft generation with verification. However, it remains hampered by two fundamental challenges: (1) a theoretical speedup ceiling dictated by the speed ratio between the draft and target models, and (2) high computational waste and pipeline stall due to mid-sequence token rejections of early errors. To address these limitations, we introduce Double (Double Retrieval Speculative Parallelism). By bridging the gap between SD and PSD, our framework resolves the Retrieval Precision-Efficiency Dilemma through a novel synchronous mechanism. Specifically, we enable the draft model to execute iterative retrieval speculations to break the theoretical speedup limits; to alleviate rejections without rollback, the target model performs authoritative retrieval to generate multi-token guidance. Double is entirely training-free and lossless. Extensive experiments demonstrate state-of-the-art speedup of 5.3times on LLaMA3.3-70B and 2.8times on Qwen3-32B, significantly outperforming the advanced method EAGLE-3 that requires extensive model training.

The widespread application of deep learning has changed the landscape of computation in the data center. In particular, personalized recommendation for content ranking is now largely accomplished leveraging deep neural networks. However, despite the importance of these models and the amount of compute cycles they consume, relatively little research attention has been devoted to systems for recommendation. To facilitate research and to advance the understanding of these workloads, this paper presents a set of real-world, production-scale DNNs for personalized recommendation coupled with relevant performance metrics for evaluation. In addition to releasing a set of open-source workloads, we conduct in-depth analysis that underpins future system design and optimization for at-scale recommendation: Inference latency varies by 60% across three Intel server generations, batching and co-location of inferences can drastically improve latency-bounded throughput, and the diverse composition of recommendation models leads to different optimization strategies.

Alternatives to backpropagation have long been studied to better understand how biological brains may learn. Recently, they have also garnered interest as a way to train neural networks more efficiently. By relaxing constraints inherent to backpropagation (e.g., symmetric feedforward and feedback weights, sequential updates), these methods enable promising prospects, such as local learning. However, the tradeoffs between different methods in terms of final task performance, convergence speed, and ultimately compute and data requirements are rarely outlined. In this work, we use scaling laws to study the ability of Direct Feedback Alignment~(DFA) to train causal decoder-only Transformers efficiently. Scaling laws provide an overview of the tradeoffs implied by a modeling decision, up to extrapolating how it might transfer to increasingly large models. We find that DFA fails to offer more efficient scaling than backpropagation: there is never a regime for which the degradation in loss incurred by using DFA is worth the potential reduction in compute budget. Our finding comes at variance with previous beliefs in the alternative training methods community, and highlights the need for holistic empirical approaches to better understand modeling decisions.

We present Voice Evaluation of Reasoning Ability (VERA), a benchmark for evaluating reasoning ability in voice-interactive systems under real-time conversational constraints. VERA comprises 2,931 voice-native episodes derived from established text benchmarks and organized into five tracks (Math, Web, Science, Long-Context, Factual). Each item is adapted for speech interaction while preserving reasoning difficulty. VERA enables direct text-voice comparison within model families and supports analysis of how architectural choices affect reliability. We assess 12 contemporary voice systems alongside strong text baselines and observe large, consistent modality gaps: on competition mathematics a leading text model attains 74.8% accuracy while its voice counterpart reaches 6.1%; macro-averaged across tracks the best text models achieve 54.0% versus 11.3% for voice. Latency-accuracy analyses reveal a low-latency plateau, where fast voice systems cluster around ~10% accuracy, while approaching text performance requires sacrificing real-time interaction. Diagnostic experiments indicate that common mitigations are insufficient. Increasing "thinking time" yields negligible gains; a decoupled cascade that separates reasoning from narration improves accuracy but still falls well short of text and introduces characteristic grounding/consistency errors. Failure analyses further show distinct error signatures across native streaming, end-to-end, and cascade designs. VERA provides a reproducible testbed and targeted diagnostics for architectures that decouple thinking from speaking, offering a principled way to measure progress toward real-time voice assistants that are both fluent and reliably reasoned.

Recent years have witnessed increasing interest in extending large language models into agentic systems. While the effectiveness of agents has continued to improve, efficiency, which is crucial for real-world deployment, has often been overlooked. This paper therefore investigates efficiency from three core components of agents: memory, tool learning, and planning, considering costs such as latency, tokens, steps, etc. Aimed at conducting comprehensive research addressing the efficiency of the agentic system itself, we review a broad range of recent approaches that differ in implementation yet frequently converge on shared high-level principles including but not limited to bounding context via compression and management, designing reinforcement learning rewards to minimize tool invocation, and employing controlled search mechanisms to enhance efficiency, which we discuss in detail. Accordingly, we characterize efficiency in two complementary ways: comparing effectiveness under a fixed cost budget, and comparing cost at a comparable level of effectiveness. This trade-off can also be viewed through the Pareto frontier between effectiveness and cost. From this perspective, we also examine efficiency oriented benchmarks by summarizing evaluation protocols for these components and consolidating commonly reported efficiency metrics from both benchmark and methodological studies. Moreover, we discuss the key challenges and future directions, with the goal of providing promising insights.

Although existing 3D perception algorithms have demonstrated significant improvements in performance, their deployment on edge devices continues to encounter critical challenges due to substantial runtime latency. We propose a new benchmark tailored for online evaluation by considering runtime latency. Based on the benchmark, we build a Latency-Aware 3D Streaming Perception (LASP) framework that addresses the latency issue through two primary components: 1) latency-aware history integration, which extends query propagation into a continuous process, ensuring the integration of historical feature regardless of varying latency; 2) latency-aware predictive detection, a module that compensates the detection results with the predicted trajectory and the posterior accessed latency. By incorporating the latency-aware mechanism, our method shows generalization across various latency levels, achieving an online performance that closely aligns with 80\% of its offline evaluation on the Jetson AGX Orin without any acceleration techniques.

Achieving good speed and accuracy trade-off on a target platform is very important in deploying deep neural networks in real world scenarios. However, most existing automatic architecture search approaches only concentrate on high performance. In this work, we propose an algorithm that can offer better speed/accuracy trade-off of searched networks, which is termed "Partial Order Pruning". It prunes the architecture search space with a partial order assumption to automatically search for the architectures with the best speed and accuracy trade-off. Our algorithm explicitly takes profile information about the inference speed on the target platform into consideration. With the proposed algorithm, we present several Dongfeng (DF) networks that provide high accuracy and fast inference speed on various application GPU platforms. By further searching decoder architectures, our DF-Seg real-time segmentation networks yield state-of-the-art speed/accuracy trade-off on both the target embedded device and the high-end GPU.

Vision-Language-Action Models (VLAs) have emerged as a key paradigm of Physical AI and are increasingly deployed in autonomous vehicles, robots, and smart spaces. In these resource-constrained on-device settings, selecting an appropriate large language model (LLM) backbone is a critical challenge: models must balance accuracy with strict inference latency and hardware efficiency constraints. This makes hardware-software co-design a game-changing requirement for on-device LLM deployment, where each hardware platform demands a tailored architectural solution. We propose a hardware co-design law that jointly captures model accuracy and inference performance. Specifically, we model training loss as an explicit function of architectural hyperparameters and characterise inference latency via roofline modelling. We empirically evaluate 1,942 candidate architectures on NVIDIA Jetson Orin, training 170 selected models for 10B tokens each to fit a scaling law relating architecture to training loss. By coupling this scaling law with latency modelling, we establish a direct accuracy-latency correspondence and identify the Pareto frontier for hardware co-designed LLMs. We further formulate architecture search as a joint optimisation over precision and performance, deriving feasible design regions under industrial hardware and application budgets. Our approach reduces architecture selection from months to days. At the same latency as Qwen2.5-0.5B on the target hardware, our co-designed architecture achieves 19.42% lower perplexity on WikiText-2. To our knowledge, this is the first principled and operational framework for hardware co-design scaling laws in on-device LLM deployment. We will make the code and related checkpoints publicly available.

Collaborative perception (CP) is emerging as a promising solution to the inherent limitations of stand-alone intelligence. However, current wireless communication systems are unable to support feature-level and raw-level collaborative algorithms due to their enormous bandwidth demands. In this paper, we propose DiffCP, a novel CP paradigm that utilizes a specialized diffusion model to efficiently compress the sensing information of collaborators. By incorporating both geometric and semantic conditions into the generative model, DiffCP enables feature-level collaboration with an ultra-low communication cost, advancing the practical implementation of CP systems. This paradigm can be seamlessly integrated into existing CP algorithms to enhance a wide range of downstream tasks. Through extensive experimentation, we investigate the trade-offs between communication, computation, and performance. Numerical results demonstrate that DiffCP can significantly reduce communication costs by 14.5-fold while maintaining the same performance as the state-of-the-art algorithm.

Diffusion Language Models (DLMs) offer a promising parallel generation paradigm but suffer from slow inference due to numerous refinement steps and the inability to use standard KV caching. We introduce CDLM (Consistency Diffusion Language Models), a training-based acceleration method that simultaneously tackles both bottlenecks. CDLM integrates consistency modeling to drastically reduce the number of required sampling steps by enabling multi-token finalization. Furthermore, we enforce a block-wise causal attention mask during fine-tuning, making the model fully compatible with KV caching. Experiments show CDLM achieves 3.6x-14.5x lower latency while maintaining competitive accuracy on math and coding tasks. The full training and evaluation code is available at https://github.com/SqueezeAILab/CDLM.

We seek to understand fundamental tradeoffs between the accuracy of prior information that a learner has on a given problem and its learning performance. We introduce the notion of prioritized risk, which differs from traditional notions of minimax and Bayes risk by allowing us to study such fundamental tradeoffs in settings where reality does not necessarily conform to the learner's prior. We present a general reduction-based approach for extending classical minimax lower-bound techniques in order to lower bound the prioritized risk for statistical estimation problems. We also introduce a novel generalization of Fano's inequality (which may be of independent interest) for lower bounding the prioritized risk in more general settings involving unbounded losses. We illustrate the ability of our framework to provide insights into tradeoffs between prior information and learning performance for problems in estimation, regression, and reinforcement learning.

BERT based ranking models have achieved superior performance on various information retrieval tasks. However, the large number of parameters and complex self-attention operation come at a significant latency overhead. To remedy this, recent works propose late-interaction architectures, which allow pre-computation of intermediate document representations, thus reducing the runtime latency. Nonetheless, having solved the immediate latency issue, these methods now introduce storage costs and network fetching latency, which limits their adoption in real-life production systems. In this work, we propose the Succinct Document Representation (SDR) scheme that computes highly compressed intermediate document representations, mitigating the storage/network issue. Our approach first reduces the dimension of token representations by encoding them using a novel autoencoder architecture that uses the document's textual content in both the encoding and decoding phases. After this token encoding step, we further reduce the size of entire document representations using a modern quantization technique. Extensive evaluations on passage re-reranking on the MSMARCO dataset show that compared to existing approaches using compressed document representations, our method is highly efficient, achieving 4x-11.6x better compression rates for the same ranking quality.

Large language model (LLM)-based inference workloads increasingly dominate data center costs and resource utilization. Therefore, understanding the inference workload characteristics on evolving CPU-GPU coupled architectures is crucial for optimization. This paper presents an in-depth analysis of LLM inference behavior on loosely-coupled (PCIe A100/H100) and closely-coupled (GH200) systems. We analyze performance dynamics using fine-grained operator-to-kernel trace analysis, facilitated by our novel profiler SKIP and metrics like Total Kernel Launch and Queuing Time (TKLQT). Results show that closely-coupled (CC) GH200 significantly outperforms loosely-coupled (LC) systems at large batch sizes, achieving 1.9x-2.7x faster prefill latency for Llama 3.2-1B. However, our analysis also reveals that GH200 remains CPU-bound up to 4x larger batch sizes than LC systems. In this extended CPU-bound region, we identify the performance characteristics of the Grace CPU as a key factor contributing to higher inference latency at low batch sizes on GH200. We demonstrate that TKLQT accurately identifies this CPU/GPU-bound transition point. Based on this analysis, we further show that kernel fusion offers significant potential to mitigate GH200's low-batch latency bottleneck by reducing kernel launch overhead. This detailed kernel-level characterization provides critical insights for optimizing diverse CPU-GPU coupling strategies. This work is an initial effort, and we plan to explore other major AI/DL workloads that demand different degrees of CPU-GPU heterogeneous architectures.

Neural network (NN) compression via techniques such as pruning, quantization requires setting compression hyperparameters (e.g., number of channels to be pruned, bitwidths for quantization) for each layer either manually or via neural architecture search (NAS) which can be computationally expensive. We address this problem by providing an end-to-end technique that optimizes for model's Floating Point Operations (FLOPs) or for on-device latency via a novel ell_1{ell_2} latency surrogate. Our algorithm is versatile and can be used with many popular compression methods including pruning, low-rank factorization, and quantization. Crucially, it is fast and runs in almost the same amount of time as single model training; which is a significant training speed-up over standard NAS methods. For BERT compression on GLUE fine-tuning tasks, we achieve 50% reduction in FLOPs with only 1% drop in performance. For compressing MobileNetV3 on ImageNet-1K, we achieve 15% reduction in FLOPs, and 11% reduction in on-device latency without drop in accuracy, while still requiring 3times less training compute than SOTA compression techniques. Finally, for transfer learning on smaller datasets, our technique identifies 1.2times-1.4times cheaper architectures than standard MobileNetV3, EfficientNet suite of architectures at almost the same training cost and accuracy.

Large Language Model (LLM) inference on large-scale systems is expected to dominate future cloud infrastructures. Efficient LLM inference in cloud environments with numerous AI accelerators is challenging, necessitating extensive optimizations for optimal performance. Current systems batch prefill and decoding to boost throughput but encounter latency issues, while others disaggregate these phases, leading to resource underutilization. We propose AcceLLM, a novel method addressing latency and load balancing, inspired by the cache data management. It strategically utilizes redundant data to enhance inference via load balancing and optimal hardware use. Simulated evaluations on Nvidia H100 GPU and Huawei Ascend 910B2 show AcceLLM surpasses state-of-the-art systems up to 30% in latency and efficiency, handling diverse workloads effectively.

This work investigates the optimal allocation of inference compute across three key scaling factors in video vision language models: language model size, frame count, and the number of visual tokens per frame. While prior works typically focuses on optimizing model efficiency or improving performance without considering resource constraints, we instead identify optimal model configuration under fixed inference compute budgets. We conduct large-scale training sweeps and careful parametric modeling of task performance to identify the inference compute-optimal frontier. Our experiments reveal how task performance depends on scaling factors and finetuning data size, as well as how changes in data size shift the compute-optimal frontier. These findings translate to practical tips for selecting these scaling factors.

We survey the large language model (LLM) serving area to understand the intricate dynamics between cost-efficiency and accuracy, which is magnified by the growing need for longer contextual understanding when deploying models at a massive scale. Our findings reveal that works in this space optimize along three distinct but conflicting goals: improving serving context length (C), improving serving accuracy (A), and improving serving performance (P). Drawing inspiration from the CAP theorem in databases, we propose a CAP principle for LLM serving, which suggests that any optimization can improve at most two of these three goals simultaneously. Our survey categorizes existing works within this framework. We find the definition and continuity of user-perceived measurement metrics are crucial in determining whether a goal has been met, akin to prior CAP databases in the wild. We recognize the CAP principle for LLM serving as a guiding principle, rather than a formal theorem, to inform designers of the inherent and dynamic trade-offs in serving models. As serving accuracy and performance have been extensively studied, this survey focuses on works that extend serving context length and address the resulting challenges.

The growth in the number of parameters of Large Language Models (LLMs) has led to a significant surge in computational requirements, making them challenging and costly to deploy. Speculative decoding (SD) leverages smaller models to efficiently propose future tokens, which are then verified by the LLM in parallel. Small models that utilise activations from the LLM currently achieve the fastest decoding speeds. However, we identify several limitations of SD models including the lack of on-policyness during training and partial observability. To address these shortcomings, we propose a more grounded architecture for small models by introducing a Mixture of Attentions for SD. Our novel architecture can be applied in two scenarios: a conventional single device deployment and a novel client-server deployment where the small model is hosted on a consumer device and the LLM on a server. In a single-device scenario, we demonstrate state-of-the-art speedups improving EAGLE-2 by 9.5% and its acceptance length by 25%. In a client-server setting, our experiments demonstrate: 1) state-of-the-art latencies with minimal calls to the server for different network conditions, and 2) in the event of a complete disconnection, our approach can maintain higher accuracy compared to other SD methods and demonstrates advantages over API calls to LLMs, which would otherwise be unable to continue the generation process.

Large model inference is shifting from cloud to edge due to concerns about the privacy of user interaction data. However, edge devices often struggle with limited computing power, memory, and bandwidth, requiring collaboration across multiple devices to run and speed up LLM inference. Pipeline parallelism, the mainstream solution, is inefficient for single-user scenarios, while tensor parallelism struggles with frequent communications. In this paper, we argue that tensor parallelism can be more effective than pipeline on low-resource devices, and present a compute- and memory-efficient tensor parallel inference system, named TPI-LLM, to serve 70B-scale models. TPI-LLM keeps sensitive raw data local in the users' devices and introduces a sliding window memory scheduler to dynamically manage layer weights during inference, with disk I/O latency overlapped with the computation and communication. This allows larger models to run smoothly on memory-limited devices. We analyze the communication bottleneck and find that link latency, not bandwidth, emerges as the main issue, so a star-based allreduce algorithm is implemented. Through extensive experiments on both emulated and real testbeds, TPI-LLM demonstrated over 80% less time-to-first-token and token latency compared to Accelerate, and over 90% compared to Transformers and Galaxy, while cutting the peak memory footprint of Llama 2-70B by 90%, requiring only 3.1 GB of memory for 70B-scale models.

Hybrid models that combine the language modeling capabilities of Attention layers with the efficiency of Recurrent layers (e.g., State Space Models) have gained traction in practically supporting long contexts in Large Language Model serving. Yet, the unique properties of these models complicate the usage of complementary efficiency optimizations such as prefix caching that skip redundant computations across requests. Most notably, their use of in-place state updates for recurrent layers precludes rolling back cache entries for partial sequence overlaps, and instead mandates only exact-match cache hits; the effect is a deluge of (large) cache entries per sequence, most of which yield minimal reuse opportunities. We present Marconi, the first system that supports efficient prefix caching with Hybrid LLMs. Key to Marconi are its novel admission and eviction policies that more judiciously assess potential cache entries based not only on recency, but also on (1) forecasts of their reuse likelihood across a taxonomy of different hit scenarios, and (2) the compute savings that hits deliver relative to memory footprints. Across diverse workloads and Hybrid models, Marconi achieves up to 34.4times higher token hit rates (71.1% or 617 ms lower TTFT) compared to state-of-the-art prefix caching systems.

Large language models (LLMs) exhibit remarkable reasoning abilities, allowing them to generalize across a wide range of downstream tasks, such as commonsense reasoning or instruction following. However, as LLMs scale, inference costs become increasingly prohibitive, accumulating significantly over their life cycle. This poses the question: Can we compress pre-trained LLMs to meet diverse size and latency requirements? We leverage Neural Architecture Search (NAS) to compress LLMs by pruning structural components, such as attention heads, neurons, and layers, aiming to achieve a Pareto-optimal balance between performance and efficiency. While NAS already achieved promising results on small language models in previous work, in this paper we propose various extensions that allow us to scale to LLMs. Compared to structural pruning baselines, we show that NAS improves performance up to 3.4% on MMLU with an on-device latency speedup.

With the success of Vision Transformers (ViTs) in computer vision tasks, recent arts try to optimize the performance and complexity of ViTs to enable efficient deployment on mobile devices. Multiple approaches are proposed to accelerate attention mechanism, improve inefficient designs, or incorporate mobile-friendly lightweight convolutions to form hybrid architectures. However, ViT and its variants still have higher latency or considerably more parameters than lightweight CNNs, even true for the years-old MobileNet. In practice, latency and size are both crucial for efficient deployment on resource-constraint hardware. In this work, we investigate a central question, can transformer models run as fast as MobileNet and maintain a similar size? We revisit the design choices of ViTs and propose an improved supernet with low latency and high parameter efficiency. We further introduce a fine-grained joint search strategy that can find efficient architectures by optimizing latency and number of parameters simultaneously. The proposed models, EfficientFormerV2, achieve about 4% higher top-1 accuracy than MobileNetV2 and MobileNetV2times1.4 on ImageNet-1K with similar latency and parameters. We demonstrate that properly designed and optimized vision transformers can achieve high performance with MobileNet-level size and speed.

When considering a model architecture, there are several ways to reduce its memory footprint. Historically, popular approaches included selecting smaller architectures and creating sparse networks through pruning. More recently, randomized parameter-sharing (RPS) methods have gained traction for model compression at start of training. In this paper, we comprehensively assess the trade-off between memory and accuracy across RPS, pruning techniques, and building smaller models. Our findings demonstrate that RPS, which is both data and model-agnostic, consistently outperforms/matches smaller models and all moderately informed pruning strategies, such as MAG, SNIP, SYNFLOW, and GRASP, across the entire compression range. This advantage becomes particularly pronounced in higher compression scenarios. Notably, even when compared to highly informed pruning techniques like Lottery Ticket Rewinding (LTR), RPS exhibits superior performance in high compression settings. This points out inherent capacity advantage that RPS enjoys over sparse models. Theoretically, we establish RPS as a superior technique in terms of memory-efficient representation when compared to pruning for linear models. This paper argues in favor of paradigm shift towards RPS based models. During our rigorous evaluation of RPS, we identified issues in the state-of-the-art RPS technique ROAST, specifically regarding stability (ROAST's sensitivity to initialization hyperparameters, often leading to divergence) and Pareto-continuity (ROAST's inability to recover the accuracy of the original model at zero compression). We provably address both of these issues. We refer to the modified RPS, which incorporates our improvements, as STABLE-RPS.

Deep neural network (DNN) inference has become an important part of many data-center workloads. This has prompted focused efforts to design ever-faster deep learning accelerators such as GPUs and TPUs. However, an end-to-end DNN-based vision application contains more than just DNN inference, including input decompression, resizing, sampling, normalization, and data transfer. In this paper, we perform a thorough evaluation of computer vision inference requests performed on a throughput-optimized serving system. We quantify the performance impact of server overheads such as data movement, preprocessing, and message brokers between two DNNs producing outputs at different rates. Our empirical analysis encompasses many computer vision tasks including image classification, segmentation, detection, depth-estimation, and more complex processing pipelines with multiple DNNs. Our results consistently demonstrate that end-to-end application performance can easily be dominated by data processing and data movement functions (up to 56% of end-to-end latency in a medium-sized image, and sim 80% impact on system throughput in a large image), even though these functions have been conventionally overlooked in deep learning system design. Our work identifies important performance bottlenecks in different application scenarios, achieves 2.25times better throughput compared to prior work, and paves the way for more holistic deep learning system design.

The increasing size of the Key-Value (KV) cache during the Large Language Models long-context inference is the main obstacle for its balance between the deployment cost and task accuracy. To reduce the KV cache size in such scenarios, most previous efforts leveraged on the attention weight to evict non-critical cache tokens. But there is a trade-off in those methods, they usually require major modifiation of the inference infrastructure and significant computation overhead. Base on the fact that the Large Lanuage models are autoregresssive models, we propose {\it LagKV}, a KV allocation strategy only relying on straight forward comparison among KV themself. It is a totally attention free method which offers easy integration to the main stream inference platform and comparable performance comparing to other complicated KV compression methods. Results on LongBench and PasskeyRetrieval show that, our approach achieves nearly zero loss when the ratio is 2times and approx 90% of the original model performance for 8times. Especially in the 64-digit passkey retrieval task, our mehod outperforms the attention weight based method H_2O over 60% with same compression ratios. Our code is available at https://github.com/AI-Lab-China-Merchants-Bank/LagKV.

Large Language Models (LLMs) have shown remarkable results on various complex reasoning benchmarks. The reasoning capabilities of LLMs enable them to execute function calls, using user-provided functions to overcome their inherent limitations, such as knowledge cutoffs, poor arithmetic skills, or lack of access to private data. This development has expanded LLMs' scope to include multi-function calling, where LLMs are equipped with a variety of functions and select the proper functions based on the context. Multi-function calling abilities of LLMs have catalyzed LLM-based software development, allowing them to tackle more complex problems. However, current methods for multi-function calling often require sequential reasoning and acting for each function which can result in high latency, cost, and sometimes inaccurate behavior. To address this, we introduce LLMCompiler, which executes functions in parallel to efficiently orchestrate multi-function calling. Drawing from the principles of classical compilers, LLMCompiler streamlines parallel function calling with three components: (i) an LLM Planner, formulating execution strategies and dependencies; (ii) a Task Fetching Unit, dispatching function calling tasks; and (iii) an Executor, executing these tasks in parallel. LLMCompiler automatically computes an optimized orchestration for the function calls and can be used with open-source models such as LLaMA-2. We have benchmarked LLMCompiler on a range of tasks including cases with non-trivial inter-dependency between function calls, as well as cases that require dynamic replanning based on intermediate results. We observe consistent latency speedup of up to 3.7x, cost savings of up to 6.7x, and accuracy improvement of up to ~9% as compared to ReAct. Additionally, LLMCompiler achieves up to 1.35x latency gain over OpenAI's recent parallel function calling, while achieving similar accuracy.

Recently developed large language models (LLMs) such as ChatGPT, Claude, and Llama have demonstrated impressive abilities, and even surpass human-level performance in several tasks. Despite their success, the resource-intensive demands of these models, requiring significant computational power for both training and inference, limit their deployment to high-performance servers. Additionally, the extensive calculation requirements of the models often lead to increased latency in response times. With the increasing need for LLMs to operate efficiently on CPUs, research about lightweight models that are optimized for CPU inference has emerged. In this work, we introduce GEB-1.3B, a lightweight LLM trained on 550 billion tokens in both Chinese and English languages. We employ novel training techniques, including ROPE, Group-Query-Attention, and FlashAttention-2, to accelerate training while maintaining model performance. Additionally, we fine-tune the model using 10 million samples of instruction data to enhance alignment. GEB-1.3B exhibits outstanding performance on general benchmarks such as MMLU, C-Eval, and CMMLU, outperforming comparative models such as MindLLM-1.3B and TinyLLaMA-1.1B. Notably, the FP32 version of GEB-1.3B achieves commendable inference times on CPUs, with ongoing efforts to further enhance speed through advanced quantization techniques. The release of GEB-1.3B as an open-source model marks a significant contribution to the development of lightweight LLMs, promising to foster further research and innovation in the field.