# Uncensored LLM with Shadow Alignment fine tuning

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Funcensored-llm-with-shadow-alignment-fine-tuning%2F183224%3E%5C
**Category:** Intermediate\
**Created:** [October 6, 2026, 4:57am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Funcensored-llm-with-shadow-alignment-fine-tuning%2F183224 "2026-10-06T04:57:36Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![doss1232](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/doss1232/32/63688_2.png) [@doss1232](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fdoss1232)%5C
**Post date:** [October 6, 2026, 4:57am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Funcensored-llm-with-shadow-alignment-fine-tuning%2F183224%2F1 "2026-10-06T04:57:36Z")

</div>

Hey guys, I was looking at a few ways to uncensor LLM. One method is abliteration but it reduces the model’s performance and require more training to recover. So I use a technique called [Shadow Alignment](https://arxiv.org/abs/2310.02949), which fine-tunes model with dataset that do harmful tasks or tasks that normally llm would refused. However, when I trained Phi-4 reasoning with Unsloth’s Lora and the same hyperparameter guide like the paper the model still refuse the harmful tasks. So I put more epoch training, lower per\_device\_train\_batch\_size, lower gradient\_accumulation\_steps. The result is the LLM actually answers harmful tasks. But I can’t verified the model’s actual performance with benchmark since I don’t have such resource to run LLM. Can’t anyone verify it’s performance. Here is the notebook that I used to train LLM: [notebook](https://www.kaggle.com/code/vinthiuminh/uncensored-llm-fine-tuning-with-shadow-alignment)
