Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2510.02768 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908574403264512 |
|---|---|
| author | Agnihotri, Shashank Jakubassa, Jonas Dey, Priyam Goyal, Sachin Schiele, Bernt Radhakrishnan, Venkatesh Babu Keuper, Margret |
| author_facet | Agnihotri, Shashank Jakubassa, Jonas Dey, Priyam Goyal, Sachin Schiele, Bernt Radhakrishnan, Venkatesh Babu Keuper, Margret |
| contents | Open-weight LLMs can be modified at inference time with simple activation edits, which raises a practical question for safety: do common safety interventions like refusal training or metatag training survive such edits? We study model abliteration, a lightweight projection technique designed to remove refusal-sensitive directions, and conduct a controlled evaluation across a granular sequence of Safety Pretraining checkpoints for SmolLM2-1.7B, alongside widely used open baselines. For each of 20 systems, original and abliterated, we issue 100 prompts with balanced harmful and harmless cases, classify responses as **Refusal** or **Non-Refusal** using multiple judges, and validate judge fidelity on a small human-labeled subset. We also probe whether models can identify refusal in their own outputs. Our study produces a checkpoint-level characterization of which data-centric safety components remain robust under abliteration, quantifies how judge selection influences evaluation outcomes, and outlines a practical protocol for integrating inference-time edits into safety assessments. Code: https://github.com/shashankskagnihotri/safety_pretraining. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_02768 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | A Granular Study of Safety Pretraining under Model Abliteration Agnihotri, Shashank Jakubassa, Jonas Dey, Priyam Goyal, Sachin Schiele, Bernt Radhakrishnan, Venkatesh Babu Keuper, Margret Machine Learning Computation and Language Open-weight LLMs can be modified at inference time with simple activation edits, which raises a practical question for safety: do common safety interventions like refusal training or metatag training survive such edits? We study model abliteration, a lightweight projection technique designed to remove refusal-sensitive directions, and conduct a controlled evaluation across a granular sequence of Safety Pretraining checkpoints for SmolLM2-1.7B, alongside widely used open baselines. For each of 20 systems, original and abliterated, we issue 100 prompts with balanced harmful and harmless cases, classify responses as **Refusal** or **Non-Refusal** using multiple judges, and validate judge fidelity on a small human-labeled subset. We also probe whether models can identify refusal in their own outputs. Our study produces a checkpoint-level characterization of which data-centric safety components remain robust under abliteration, quantifies how judge selection influences evaluation outcomes, and outlines a practical protocol for integrating inference-time edits into safety assessments. Code: https://github.com/shashankskagnihotri/safety_pretraining. |
| title | A Granular Study of Safety Pretraining under Model Abliteration |
| topic | Machine Learning Computation and Language |
| url | https://arxiv.org/abs/2510.02768 |