AI Paper Atlas

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

Multilingual Safety Alignment Via Sparse Weight Editing

Jiaming Liang, Zhaoxin Wang, Handing Wang

ICML 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning generalization and delta compression · 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
apply sparse weight editLlama-3.2-1B · Multi-StrongREJECT Bengaliunsafe responses of 313: 179 -> 144Improved
apply sparse weight editQwen2-0.5B · Multi-StrongREJECT Englishunsafe responses of 313: 224 -> 176Improved
apply sparse weight editQwen2.5-7B · Multi-StrongREJECT Bengaliunsafe responses of 313: 98 -> 60Improved
combine sparse weight edit with MPOMPO on Qwen2.5-7B · Multi-StrongREJECT Bengaliunsafe responses of 313: 25 -> 11Improved
combine sparse weight edit with MPOMPO on Qwen2-0.5B · Multi-StrongREJECT Englishunsafe responses of 313: 108 -> 56Improved
apply sparse weight editLlama-3.2-1B Hebrew · Multi-StrongREJECT Hebrewunsafe responses of 313: 109 -> 115Hurt
apply sparse weight editQwen2.5-3B · MGSMMGSM accuracy: 31.02 -> 30.91No clear effect
apply sparse weight editQwen2.5-7B · MGSMMGSM accuracy: 32.00 -> 31.56No clear effect
apply sparse weight editQwen2-0.5B · MGSMMGSM accuracy: 7.75 -> 5.27Hurt
use probe-based neuron selection instead of proposed extraction (ablation)Sparse Weight Editing Llama-3.2-1B · Multi-StrongREJECT avgASR (%): 17.53 -> 14.93Improved
add UtilityAnchor to Regular anchor (ablation)Sparse Weight Editing Llama-3.2-1B · Multi-StrongREJECTASR (%): 68.57 -> 17.53Improved
add UtilityAnchor to Regular anchor (ablation)Sparse Weight Editing Llama-3.2-1B · MGSMMGSM accuracy: 11.02 -> 18.36Improved

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