paper-with-me

홈 › Papers

Uncovering Cross-Linguistic Disparities in LLMs using Sparse Autoencoders

2025-07-25 · Richmond Sin Jing Xuan, Jalil Huseynov, Yang Zhang arxiv

Multilingual large language models (LLMs) exhibit strong cross-linguistic generalization, yet medium to low resource languages underperform on common benchmarks such as ARC-Challenge, MMLU, and HellaSwag. We analyze activation patterns in Gemma-2-2B across all 26 residual layers and 10 languages: Chinese (zh), Russian (ru), Spanish (es), Italian (it), medium to low resource languages including Indonesian (id), Catalan (ca), Marathi (mr), Malayalam (ml), and Hindi (hi), with English (en) as the reference. Using Sparse Autoencoders (SAEs), we reveal systematic disparities in activation patterns. Medium to low resource languages receive up to 26.27 percent lower activations in early layers, with a persistent gap of 19.89 percent in deeper layers. To address this, we apply activation-aware fine-tuning via Low-Rank Adaptation (LoRA), leading to substantial activation gains, such as 87.69 percent for Malayalam and 86.32 percent for Hindi, while maintaining English retention at approximately 91 percent. After fine-tuning, benchmark results show modest but consistent improvements, highlighting activation alignment as a key factor in enhancing multilingual LLM performance.

📄 PDF Abstract BibTeX arXiv:2507.18918

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Uncovering inequalities in new knowledge learning by large language models across different languages

2025-03-06 · Chenglong Wang, Haoyu Tang, Xiyuan Yang, Yueqi Xie 외

As large language models (LLMs) gradually become integral tools for problem solving in daily life worldwide, understanding linguistic inequality is becoming increasingly important. Existing research has primarily focused…

In-Context Learning

One Model, Many Morals: Uncovering Cross-Linguistic Misalignments in Computational Moral Reasoning

2025-09-25 · Sualeha Farid, Jayden Lin, Zean Chen, Shivani Kumar 외 arxiv

Large Language Models (LLMs) are increasingly deployed in multilingual and multicultural environments where moral reasoning is essential for generating ethically appropriate responses. Yet, the dominant pretraining of LL…

Actions Speak Louder than Words: Agent Decisions Reveal Implicit Biases in Language Models

2025-01-29 · YuXuan Li, Hirokazu Shirado, Sauvik Das

While advances in fairness and alignment have helped mitigate overt biases exhibited by large language models (LLMs) when explicitly prompted, we hypothesize that these models may still exhibit implicit biases when simul…

Decision MakingFairness

The World According to LLMs: How Geographic Origin Influences LLMs' Entity Deduction Capabilities

2025-08-07 · Harsh Nishant Lalai, Raj Sanjay Shah, Jiaxin Pei, Sashank Varma 외 arxiv

Large Language Models (LLMs) have been extensively tuned to mitigate explicit biases, yet they often exhibit subtle implicit biases rooted in their pre-training data. Rather than directly probing LLMs with human-crafted …

Tokenization Disparities as Infrastructure Bias: How Subword Systems Create Inequities in LLM Access and Efficiency

2025-10-14 · Hailay Kidu Teklehaymanot, Wolfgang Nejdl arxiv

Tokenization disparities pose a significant barrier to achieving equitable access to artificial intelligence across linguistically diverse populations. This study conducts a large-scale cross-linguistic evaluation of tok…