AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Bridging the Language Gap: Uncovering and Aligning Shared Circuits for Multi-Hop Reasoning in Multilingual LLMs

Chenghao Sun, Zhen Huang, Yonggang Zhang, Xinmei Tian, Xu Shen, Jieping Ye

AAAI 2026 · Large language models › Retrieval and knowledge · approach: Knowledge editing in LLMs · 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 only the top-40 critical attention headsQwen2.5-7B-Instruct · XHopReason (8-language average, trained on Japanese)accuracy (%): 15.34 -> 18.64Improved
fine-tune all parameters instead of critical heads (ablation)Qwen2.5-7B-Instruct · XHopReason (8-language average)accuracy (%): 15.34 -> 15.17No clear effect
apply DPO instead of critical-head tuning (ablation)Qwen2.5-7B-Instruct · XHopReason (8-language average)accuracy (%): 15.34 -> 15.98No clear effect
tune 100 least critical heads rather than top critical heads (ablation)head tuning · XHopReason (8-language average)accuracy (%): 18.64 -> 17.58Hurt
tune top-20 heads instead of top-40 (ablation)head tuning · XHopReason (8-language average)accuracy (%): 18.64 -> 18.16Hurt
tune top-80 or top-160 heads instead of top-40 (ablation)head tuning · XHopReason (8-language average)accuracy (%): 18.64 -> 18.39 (top-80), 18.01 (top-160)Hurt
train head tuning on a single source language, test on other languagesTop-40 heads MFT · XHopReason cross-lingual transferaccuracy (%): Italian 19.96 -> 22.05 when trained on JapaneseImproved
fine-tune all parameters instead of critical heads (ablation)Qwen2.5-7B-Instruct · General benchmarks (MMLU, StrategyQA, INCLUDE, Belebele)average change (points): -24.58 average changeHurt
apply DPO instead of critical-head tuning (ablation)Qwen2.5-7B-Instruct · General benchmarksaverage change (points): -22.72 average changeHurt
fine-tune only top-40 headsQwen2.5-7B-Instruct · General benchmarksaverage change (points): -0.85 average changeNo clear effect
fine-tune only top-40 headsQwen2.5-7B-Instruct · StrategyQA (en)accuracy (%): 56.83 -> 61.13Improved
fine-tune only top-40 headsQwen2.5-7B-Instruct · Japanese test accuracyaccuracy (%): Top-40 20.28 vs Full Fine-tuning 22.35 and DPO 24.10Mixed

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