AI Paper Atlas

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

Toward Robust Multilingual Adaptation of LLMs for Low-Resource Languages

Haolin Li, Haipeng Zhang, Mang Li, Yaohua Wang, Lijie Wen, zhang yu, Biqing Huang

ICML 2026 · Large language models › Language, safety and interpretability · approach: Multilingual LLM data and culture

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 the LLM critic (ablation)LiRA · LazRetrievalnDCG@10: 71.29 -> 77.71Improved
add the Embeds critic (ablation)LiRA · LazRetrievalnDCG@10: 65.77 -> 77.71Improved
add the Embeds critic (ablation)LiRA · STS22Pearson: 61.78 -> 75.00Improved
add translations (ablation)LiRA · LazRetrievalnDCG@10: 75.48 -> 77.71Improved
add multilingual encoder (ablation)LiRA · STS22Pearson: 72.43 -> 75.00Improved
add the FIFO queue (ablation)LiRA · LazRetrievalnDCG@10: 64.29 -> 77.71Improved
add the LLM critic (ablation)LiRA · MGSMaccuracy (%): 68.9 -> 71.1Improved
add LiRA-Large to Qwen3-Embedding-8BQwen3-E-8B · LazRetrievalnDCG@10 average: 68.05 -> 72.86Improved
scale LiRA from Base to Large to MaxQwen3-E-8B · LazRetrievalnDCG@10 average: 69.61 -> 72.86 -> 77.71Improved
add LiRA to Qwen3-Embedding-8BQwen3-E-8B · Public retrievalaverage score: 79.57 -> 81.35Improved
add LiRA-Large to Qwen3-8BQwen3-8B · MGSMaccuracy (%): 69.4 -> 71.1Improved
add LiRA-Large to Qwen3-8BQwen3-8B · X-CSQAaccuracy (%): 62.9 -> 65.5Improved

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