# Defending Against Harmful Supervision Hidden in Benign Samples

arXiv: https://arxiv.org/abs/2606.30263 · HTML: https://mlsanity.com/paper/2606.30263/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Safety alignment, refusal and guardrails, one of 12 directions in LLM safety and security.
- Builds on 31 of its 44 references that are in the map.
- First posted on arXiv 29 Jun 2026.

## Placed in

- LLM safety and security / [Safety alignment, refusal and guardrails](https://mlsanity.com/safety/safety-alignment-refusal-and-guardrails/)

## Builds on (most cited first, 30 of 31)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Finetuned Language Models Are Zero-Shot Learners](https://mlsanity.com/paper/2109.01652/) (2021)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset](https://mlsanity.com/paper/2307.04657/) (2023)
- [Safe RLHF: Safe Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2310.12773/) (2023)
- [SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization](https://mlsanity.com/paper/1911.12237/) (2019)
- [Safety Alignment Should Be Made More Than Just a Few Tokens Deep](https://mlsanity.com/paper/2406.05946/) (2024)
- [Data Selection for Language Models via Importance Resampling](https://mlsanity.com/paper/2302.03169/) (2023)
- [Shadow Alignment: The Ease of Subverting Safely-Aligned Language Models](https://mlsanity.com/paper/2310.02949/) (2023)
- [Removing RLHF Protections in GPT-4 via Fine-Tuning](https://mlsanity.com/paper/2311.05553/) (2023)
- [A Survey on Trustworthy LLM Agents: Threats and Countermeasures](https://mlsanity.com/paper/2503.09648/) (2025)
- [Lisa: Lazy Safety Alignment for Large Language Models against Harmful Fine-tuning Attack](https://mlsanity.com/paper/2405.18641/) (2024)
- [Booster: Tackling Harmful Fine-tuning for Large Language Models via Attenuating Harmful Perturbation](https://mlsanity.com/paper/2409.01586/) (2024)
- [Targeted Vaccine: Safety Alignment for Large Language Models against Harmful Fine-Tuning via Layer-wise Perturbation](https://mlsanity.com/paper/2410.09760/) (2024)
- [Towards Understanding the Fragility of Multilingual LLMs against Fine-Tuning Attacks](https://mlsanity.com/paper/2410.18210/) (2024)
- [Virus: Harmful Fine-tuning Attack for Large Language Models Bypassing Guardrail Moderation](https://mlsanity.com/paper/2501.17433/) (2025)
- [Layer-Aware Representation Filtering: Purifying Finetuning Data to Preserve LLM Safety Alignment](https://mlsanity.com/paper/2507.18631/) (2025)
- [Beware of Your Po! Measuring and Mitigating AI Safety Risks in Role-Play Fine-Tuning of LLMs](https://mlsanity.com/paper/2502.20968/) (2025)
- [Why LLM Safety Guardrails Collapse After Fine-tuning: A Similarity Analysis Between Alignment and Fine-tuning Datasets](https://mlsanity.com/paper/2506.05346/) (2025)
- [SequentialBreak: Large Language Models Can be Fooled by Embedding Jailbreak Prompts into Sequential Prompt Chains](https://mlsanity.com/paper/2411.06426/) (2024)
- [Vulnerability-Aware Alignment: Mitigating Uneven Forgetting in Harmful Fine-Tuning](https://mlsanity.com/paper/2506.03850/) (2025)
- [The effect of fine-tuning on language model toxicity](https://mlsanity.com/paper/2410.15821/) (2024)
- [Mitigating Backdoor Threats to Large Language Models: Advancement and Challenges](https://mlsanity.com/paper/2409.19993/) (2024)
- [Adaptive Defense against Harmful Fine-Tuning for Large Language Models via Bayesian Data Scheduler](https://mlsanity.com/paper/2510.27172/) (2025)
- [LongSafety: Enhance Safety for Long-Context LLMs](https://mlsanity.com/paper/2411.06899/) (2024)
- [Shape it Up! Restoring LLM Safety during Finetuning](https://mlsanity.com/paper/2505.17196/) (2025)
- [A Guardrail for Safety Preservation: When Safety-Sensitive Subspace Meets Harmful-Resistant Null-Space](https://mlsanity.com/paper/2510.14301/) (2025)
- [Safety at One Shot: Patching Fine-Tuned LLMs with A Single Instance](https://mlsanity.com/paper/2601.01887/) (2026)

## Abstract (arXiv)

Existing defenses are effective when harmful content is explicitly mixed into downstream fine-tuning data, but crafted samples can instead hide harmful supervision inside benign tasks. We propose Embedded Attack, where harmful QA pairs are embedded within benign training samples, and show that representative guardrails often fail to detect them at the example level. To address this, we propose Dual-Reference SFT (DR-SFT), which adapts DPO-style contrastive objective design to SFT through token-level regularization, mitigating harmful fine-tuning beyond coarse data filtering.