# Alleviating the Fear of Losing Alignment in LLM Fine-tuning

arXiv: https://arxiv.org/abs/2504.09757 · HTML: https://mlsanity.com/paper/2504.09757/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 13 Apr 2025.

## Built on by (most cited first, 9 of 9)

- [Gradient Surgery for Safe LLM Fine-Tuning](https://mlsanity.com/paper/2508.07172/) (2025)
- [Re-Emergent Misalignment: How Narrow Fine-Tuning Erodes Safety Alignment in LLMs](https://mlsanity.com/paper/2507.03662/) (2025)
- [Defending MoE LLMs against Harmful Fine-Tuning via Safety Routing Alignment](https://mlsanity.com/paper/2509.22745/) (2025)
- [Safety at One Shot: Patching Fine-Tuned LLMs with A Single Instance](https://mlsanity.com/paper/2601.01887/) (2026)
- [Understanding and Preserving Safety in Fine-Tuned LLMs](https://mlsanity.com/paper/2601.10141/) (2026)
- [Surgery: Mitigating Harmful Fine-Tuning for Large Language Models via Attention Sink](https://mlsanity.com/paper/2602.05228/) (2026)
- [Safeguarding LLM Fine-tuning via Push-Pull Distributional Alignment](https://mlsanity.com/paper/2601.07200/) (2026)
- [MCP-DPT: A Defense-Placement Taxonomy and Coverage Analysis for Model Context Protocol Security](https://mlsanity.com/paper/2604.07551/) (2026)
- [Adversarial Reframing: A Framework for Targeted Generation in Language Models](https://mlsanity.com/paper/2605.21674/) (2026)

## Abstract (arXiv)

Large language models (LLMs) have demonstrated revolutionary capabilities in understanding complex contexts and performing a wide range of tasks. However, LLMs can also answer questions that are unethical or harmful, raising concerns about their applications. To regulate LLMs' responses to such questions, a training strategy called textitalignment can help. Yet, alignment can be unexpectedly compromised when fine-tuning an LLM for downstream tasks. This paper focuses on recovering the alignment lost during fine-tuning. We observe that there are two distinct directions inherent in an aligned LLM: the textitaligned direction and the textitharmful direction. An LLM is inclined to answer questions in the aligned direction while refusing queries in the harmful direction. Therefore, we propose to recover the harmful direction of the fine-tuned model that has been compromised. Specifically, we restore a small subset of the fine-tuned model's weight parameters from the original aligned model using gradient descent. We also introduce a rollback mechanism to avoid aggressive recovery and maintain downstream task performance. Our evaluation on 125 fine-tuned LLMs demonstrates that our method can reduce their harmful rate (percentage of answering harmful questions) from 33.25% to 1.74%, without sacrificing task performance much. In contrast, the existing methods either only reduce the harmful rate to a limited extent or significantly impact the normal functionality. Our code is available at https://github.com/kangyangWHU/LLMAlignment