InternVideo3: Agentify Foundation Models with Multimodal Contextual Reasoning
Abstract
InternVideo3 enhances long-horizon multimodal tasks through Multimodal Contextual Reasoning and efficient attention mechanisms, demonstrating strong performance on video understanding benchmarks and video agent capabilities.
Recent progress in foundation models has shifted toward agentic behavior involving multi-step reasoning and tool use. However, open-source efforts largely focus on text-dominant settings, leaving long-horizon multimodal tasks underexplored. This gap is evident in video tasks requiring sustained temporal understanding and iterative interaction. We present InternVideo3, a framework enhancing these capabilities via Multimodal Contextual Reasoning (MCR). MCR treats understanding as a closed-loop process over a shared, evolving context containing observations, instructions, reasoning, tool actions, and memory. This frames long-video understanding as evidence accumulation and verification. To ensure efficiency, we introduce Multimodal Multi-head Latent Attention (M^2LA), a token-preserving reparameterization compressing KV-cache states while retaining the full token stream. Our staged training includes continued pretraining, short-to-long supervised fine-tuning, rule-based reinforcement learning, and on-policy distillation. Experiments show InternVideo3 achieves strong performance on benchmarks like Video-MME, MLVU, and EgoSchema. We further instantiate the model as a video agent with retrieval tools, demonstrating robust evidence-grounded behavior. Our results suggest that efficient context handling and closed-loop reasoning are vital for adapting open multimodal models toward long-horizon visually grounded agency.
Community
A framework for long-horizon video understanding via closed-loop contextual reasoning and efficient latent attention.
Github: https://github.com/OpenGVLab/InternVideo/tree/main/InternVideo3
Get this paper in your agent:
hf papers read 2606.12195 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 1
Datasets citing this paper 1
yanziang/InternVideo3_Dataset
Spaces citing this paper 0
No Space linking this paper