# Comparative Analysis of LLM Abliteration Methods: A Cross-Architecture Evaluation

arXiv: https://arxiv.org/abs/2512.13655 · HTML: https://mlsanity.com/paper/2512.13655/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 28 of its 35 references that are in the map.
- First posted on arXiv 15 Dec 2025.

## Placed in

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

## Builds on (most cited first, 28 of 28)

- [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)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [Distributed Representations of Words and Phrases and their Compositionality](https://mlsanity.com/paper/1310.4546/) (2013)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [Jailbroken: How Does LLM Safety Training Fail?](https://mlsanity.com/paper/2307.02483/) (2023)
- [HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal](https://mlsanity.com/paper/2402.04249/) (2024)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations](https://mlsanity.com/paper/2312.06674/) (2023)
- [Refusal in Language Models Is Mediated by a Single Direction](https://mlsanity.com/paper/2406.11717/) (2024)
- [Steering Language Models With Activation Engineering](https://mlsanity.com/paper/2308.10248/) (2023)
- [The Linear Representation Hypothesis and the Geometry of Large Language Models](https://mlsanity.com/paper/2311.03658/) (2023)
- [JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models](https://mlsanity.com/paper/2404.01318/) (2024)
- [Zephyr: Direct Distillation of LM Alignment](https://mlsanity.com/paper/2310.16944/) (2023)
- [Safety Alignment Should Be Made More Than Just a Few Tokens Deep](https://mlsanity.com/paper/2406.05946/) (2024)
- [Improving Alignment and Robustness with Circuit Breakers](https://mlsanity.com/paper/2406.04313/) (2024)
- [The Geometry of Refusal in Large Language Models: Concept Cones and Representational Independence](https://mlsanity.com/paper/2502.17420/) (2025)
- [Reinforcement Learning for LLM Post-Training: A Survey](https://mlsanity.com/paper/2407.16216/) (2024)
- [LLMs Encode Harmfulness and Refusal Separately](https://mlsanity.com/paper/2507.11878/) (2025)
- [What Makes and Breaks Safety Fine-tuning? A Mechanistic Study](https://mlsanity.com/paper/2407.10264/) (2024)
- [Refusal Direction is Universal Across Safety-Aligned Languages](https://mlsanity.com/paper/2505.17306/) (2025)
- [Steering the CensorShip: Uncovering Representation Vectors for LLM "Thought" Control](https://mlsanity.com/paper/2504.17130/) (2025)
- [An Embarrassingly Simple Defense Against LLM Abliteration Attacks](https://mlsanity.com/paper/2505.19056/) (2025)
- [SOM Directions are Better than One: Multi-Directional Refusal Suppression in Language Models](https://mlsanity.com/paper/2511.08379/) (2025)
- [A Granular Study of Safety Pretraining under Model Abliteration](https://mlsanity.com/paper/2510.02768/) (2025)

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

- [Jailbreaking Leaves a Trace: Understanding and Detecting Jailbreak Attacks from Internal Representations of Large Language Models](https://mlsanity.com/paper/2602.11495/) (2026)
- [TamperBench: Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering](https://mlsanity.com/paper/2602.06911/) (2026)
- [Open-Weight LLM Fine-Tuning Defenses are Susceptible to Simple Attacks](https://mlsanity.com/paper/2605.26526/) (2026)
- [(Mis)generalization of Helpful-only Fine-tuning](https://mlsanity.com/paper/2606.04413/) (2026)
- [Not All Refusals Are Equal: How Safety Alignment Fails Cybersecurity at Scale](https://mlsanity.com/paper/2607.02714/) (2026)
- [Decoy Direction Optimization: A Post-Hoc Defense Against LLM Abliteration](https://mlsanity.com/paper/2609.16204/) (2026)
- [Fool's Gold: Defensive Deception Against Safety-Removal Attacks on Open-Weight Models](https://mlsanity.com/paper/2608.17202/) (2026)
- [Locating and Steering Refusal Beyond Attention](https://mlsanity.com/paper/2609.04721/) (2026)
- [BRANCH: Bypassing Multi-Scanner AI Guardrails](https://mlsanity.com/paper/2610.10742/) (2026)

## Abstract (arXiv)

Safety alignment mechanisms in large language models prevent responses to harmful queries through learned refusal behavior, yet these same mechanisms impede legitimate research applications including cognitive modeling, adversarial testing, and security analysis. While abliteration techniques enable surgical removal of refusal representations through directional orthogonalization, the relative effectiveness of available implementations remains uncharacterized. This study evaluates four abliteration tools (Heretic, DECCP, ErisForge, FailSpy) across sixteen instruction-tuned models (7B-14B parameters), reporting tool compatibility on all 16 models and quantitative metrics on subsets dictated by tool support. Single-pass methods demonstrated superior capability preservation on the benchmarked subset (avg GSM8K change across three models: ErisForge -0.28 pp; DECCP -0.13 pp), while Bayesian-optimized abliteration produced variable distribution shift (KL divergence: 0.043-1.646) with model-dependent capability impact. These findings provide researchers with evidence-based selection criteria for abliteration tool deployment across diverse model architectures. The principal finding indicates that mathematical reasoning capabilities exhibit the highest sensitivity to abliteration interventions, with GSM8K change ranging from +1.51 pp to -18.81 pp (-26.5% relative) depending on tool selection and model architecture.