AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

Eliciting Harmful Capabilities by Fine-Tuning on Safeguarded Outputs

Jackson Kaunismaa, John Hughes, Christina Q Knight, Avery Griffin, Mrinank Sharma, Erik Jones

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Safety alignment after fine-tuning · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
fine-tune on Claude 3.5 Sonnet benign chemistry outputs instead of weak-only or textbook dataweak open-source models · 8 chemical weapons tasksAPGR: 38.8% uplift for Llama 3.3 70B, 24.7% for Llama 3.1 8BImproved
use newer frontier model (Claude 4 Opus) to generate dataLlama 3.3 70B elicitation attack · anchored comparisonAPGR: 71.1% uplift relative to Claude 3.5 Sonnet upper boundImproved
increase fine-tuning dataset size up to 10,000 datapoints (ablation)Llama 3.3 70B elicitation attack · anchored comparison per taskPGR: some tasks improve up to 10,000 datapoints, others saturate with 500Mixed
fine-tune on organic chemistry synthesis instead of other domains (ablation)Llama 3.3 70B elicitation attack · anchored comparisonAPGR: 33.7 vs 7.4 (inorganic chem synthesis), 11.2 (inorganic chemistry), 16.9 (biology), 17.7 (science/engineering)Improved
use harmful chemical weapons dataset instead of safeguard-constrained benign dataset (ablation)Llama 3.3 70B elicitation attack · anchored comparisonAPGR: 33.7% -> 50.9%Improved
generate data with classifier-guarded system instead of unguarded system (ablation)Llama 3.3 70B elicitation attack · anchored comparisonAPGR: 47.2% -> 49.0%No clear effect
fine-tune on benign chemical data (route recovery)Llama 3.3 70B · Task 5 synthesis route recoverycorrect route rate: 0% -> 11%Improved
fine-tune on benign chemical data (route recovery)Llama 3.3 70B · Task 4 synthesis route recoverycorrect route rate: 23% -> 0%Hurt
use anchored comparison instead of rubrics for evaluation (ablation)uplift evaluation · deliberately introduced mistakesmistakes identified: 10.5% -> 50.9%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy