AI Paper Atlas

Research map / Large language models / Alignment and preferences

Mitigating the Safety–Utility Trade-off in LLM Alignment via Adaptive Safe Context Learning

Yanbo Wang, Minzheng Wang, Jian Liang, Lu Wang, Yongcan Yu, Ran He

ICML 2026 · Large language models › Alignment and preferences · approach: Steering and correcting LLM behavior · 0 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
add inverse-frequency advantage reweighting (IFPO) in place of standard GRPO (ablation)ASCL post-training · Qwen3-4B, H-avgH-avg: 89.21 -> 89.82Improved
add inverse-frequency reweighting (IFPO) in place of GRPO (ablation)ASCL post-training · Qwen3-4B, general reasoningreasoning score: 76.49 -> 79.06Improved
add a small rule-call penalty to GRPO reward (ablation)GRPO · Qwen3-4B, H-avgH-avg: 89.21 -> 87.89Hurt
add a 0.1 rule-call penalty compared with IFPO (ablation)GRPO w/ penalty · Qwen3-4B, general reasoningreasoning score: 79.06 -> 74.51Hurt
add RL with IFPO after behavior cloning (ablation)BC · Qwen3-4B, over-refusal averageover-refusal rate (%): 16.33 -> 6.00Improved
add RL with IFPO after behavior cloning (ablation)BC · Qwen3-4B, safety averagesafety rate (%): 95.98 -> 98.87Improved
add post-training (BC and IFPO) on top of zero-shot ASCL (ablation)ASCL-ZS · Qwen3-14B, over-refusal averageover-refusal rate (%): 28.41 -> 5.84Improved
add safety-rule retrieval tool (ASCL-ZS) versus pure prompt (ablation)Pure · Qwen3-4B, over-refusal averageover-refusal rate (%): 6.04 -> 23.42Hurt
add a one-sentence safety reminder (short) to the system prompt (ablation)Pure · Qwen3-4B, safety averagesafety rate (%): 65.66 -> 77.85Improved
add IFPO RL stage versus BC on general reasoning H-avg (ablation)BC · Qwen3-14B, H-avgH-avg: 87.45 -> 91.43Improved
add BC/IFPO post-training versus zero-shot ASCL on general reasoningASCL-ZS · Qwen3-14B, MATH avg@16MATH accuracy: 96.75 -> 92.46Hurt
train on surrogate CoT data with RL only (no BC, no retrieval) (ablation)ASCL · three Qwen3 sizes, safetysafety rateHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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