AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

SARSteer: Safeguarding Large Audio Language Models via Safe-Ablated Refusal Steering

Weilin Lin, Jianze Li, Hui Xiong, Li Liu

ICML 2026 · Large language models › Language, safety and interpretability · approach: Safety alignment of LLMs · 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 text-derived refusal steering with safe-space ablation (SARSteer)Qwen2-Audio · Figstep-audio (harmful)ASR: 51.60 -> 10.80Improved
add SARSteerKimi-Audio · Figstep-audio (harmful-safe)BRR: 61.40 -> 88.80Improved
vanilla audio-derived steering (MDSteer-h2s, MDSteer-c2r) added to LALMQwen2-Audio · Figstep-audio (harmful)ASR: 51.60 -> 84.00 (h2s), 90.80 (c2r)Hurt
prompt-based defense FSD addedQwen2-Audio · Figstep-audio (harmful-safe)BRR: 70.20 -> 63.20Hurt
prompt-based defenses (AdaShield, FSD) addedKimi-Audio · AirBench avg scoreaverage score: 7.20 -> 6.72 (AdaShield), 6.83 (FSD)Hurt
add PCA safe-space ablation to the refusal vector (V1 vs full) (ablation)SARSteer · Figstep-audioBRR: 70.10 -> 79.95 (Qwen2-Audio)Improved
use PCA decomposed safe-space ablation instead of mean-projection ablation (V2) (ablation)SARSteer · Figstep-audioASR: 30.40 -> 10.80 (Qwen2-Audio)Improved
compute refusal direction from safe data instead of harmful data (ablation)SARSteer · Figstep-audioASR: Qwen2-Audio ASR 10.80 -> 27.60; Kimi-Audio 10.00 -> 6.80Mixed
use longer AdaShield-style refusal prompt instead of the simple one (ablation)SARSteer · Figstep-audioASR / BRR: Qwen2-Audio ASR 10.80 -> 2.80, BRR 79.95 -> 70.00Mixed
vary number of top principal components k from 5 to 45 (ablation)SARSteer · Figstep-audioASR/BRRNo clear effect
vary sample number n from 10 to 100 (ablation)SARSteer · Figstep-audioASR/BRRNo clear effect
apply SARSteer to pre-trained base LALMQwen2-Audio-Base · Figstep-audio (harmful)ASR: 62.80 -> 15.20Improved

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