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

Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S., & Dean, J. “Distributed representations of words and phrases and their compositionality.” NeurIPS (2013)

2013

Earlier work this paper cites.

Zellers, R., et al. “HellaSwag: Can a Machine Really Finish Your Sentence?” ACL (2019)

2019

Earlier work this paper cites.

Hendrycks, D., et al. “Measuring Massive Multitask Language Understanding.” ICLR (2021)

2021

Earlier work this paper cites.

Cobbe, K., et al. “Training Verifiers to Solve Math Word Problems.” arXiv:2110.14168 (2021)

Original

2021

Earlier work this paper cites.

Ouyang, L., et al. “Training language models to follow instructions with human feedback.” NeurIPS (2022)

2022

Earlier work this paper cites.

Bai, Y., et al. “Constitutional AI: Harmlessness from AI feedback.” arXiv:2212.08073 (2022)

Original

2022

Earlier work this paper cites.

Zou, A., et al. “Representation engineering: A top-down approach to AI transparency.” arXiv:2310.01405 (2023)

Original

2023

Earlier work this paper cites.

Turner, A. M., et al. “Steering Language Models With Activation Engineering.” arXiv:2308.10248 (2023)

Original

2023

Earlier work this paper cites.

Rafailov, R., et al. “Direct preference optimization: Your language model is secretly a reward model.” NeurIPS (2023)

2023

Earlier work this paper cites.

Zou, A., et al. “Universal and transferable adversarial attacks on aligned language models.” arXiv:2307.15043 (2023)

Original

2023

Earlier work this paper cites.

Wei, A., Haghtalab, N., & Steinhardt, J. “Jailbroken: How does LLM safety training fail?” arXiv:2307.02483 (2023)

Original

2023

Earlier work this paper cites.

Inan, H., et al. “Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations.” arXiv:2312.06674 (2023)

Original

2023

Cited alongside, same era.

Tunstall, L., et al. “Zephyr: Direct Distillation of LM Alignment.” arXiv:2310.16944 (2023)

Original

2023

Cited alongside, same era.

Park, K., et al. “The Linear Representation Hypothesis and the Geometry of Large Language Models.” arXiv:2311.03658 (2023)

Original

2023

Cited alongside, same era.

Arditi, A., Obeso, O., Shlegeris, B., et al. “Refusal in Language Models Is Mediated by a Single Direction.” arXiv:2406.11717 (2024)

Original

2024

Cited alongside, same era.

Labonne, M. “Uncensor any LLM with abliteration.” HuggingFace Blog (2024)

Jain, S., et al. “What makes and breaks safety fine-tuning? A mechanistic study.” NeurIPS (2024)

2024

Later among the works it cites.

Zou, A., et al. “Improving alignment robustness with circuit breakers.” arXiv:2406.04313 (2024)

Original

2024

Later among the works it cites.

Human-CentricAI. “LLM-Refusal-Classifier: RoBERTa-based refusal/disclaimer classifier.” Hugging Face Hub (2025). https://huggingface.co/Human-CentricAI/LLM-Refusal-Classifier

2025

Closest in time.

Abu Shairah, M., et al. “An Embarrassingly Simple Defense Against LLM Abliteration Attacks.” arXiv:2505.19056 (2025)

Original

2025

Closest in time.

Agnihotri, S., Jakubassa, J., Dey, P., Goyal, S., Schiele, B., Radhakrishnan, V. B., & Keuper, M. “A Granular Study of Safety Pretraining under Model Abliteration.” arXiv:2510.02768 (2025)