Saved in:
Bibliographic Details
Main Authors: Menschikov, Mikhail, Kharitonov, Alexander, Kotyga, Maiia, Porvatov, Vadim, Zhukovskaya, Anna, Kagramanyan, David, Shvetsov, Egor, Burnaev, Evgeny
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.16134
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914196293156864
author Menschikov, Mikhail
Kharitonov, Alexander
Kotyga, Maiia
Porvatov, Vadim
Zhukovskaya, Anna
Kagramanyan, David
Shvetsov, Egor
Burnaev, Evgeny
author_facet Menschikov, Mikhail
Kharitonov, Alexander
Kotyga, Maiia
Porvatov, Vadim
Zhukovskaya, Anna
Kagramanyan, David
Shvetsov, Egor
Burnaev, Evgeny
contents Large Language Models (LLMs) exhibit position bias systematically underweighting information based on its location in the context but how this bias varies across languages and models remains unclear. We conduct a multilingual study across five typologically diverse languages (English, Russian, German, Hindi, Vietnamese) and five model architectures, analyzing how position bias interacts with prompting strategies and affects output entropy. Our key findings are: (1) Position bias is primarily model-driven but shows language-specific nuances. Notably, Qwen2.5-7B-Instruct, DeepSeek 7B Chat and Mistral 7B consistently favor late positions challenging the common assumption of universal early-token preference. (2) Explicitly instructing the model, in the presence of irrelevant distractors, that "the most relevant context to the query is marked as 1" unexpectedly reduces accuracy across all languages, questioning standard prompt-engineering practices. (3) Accuracy consistently drops most when relevant information appears in the middle of the context, yet this is not reflected in a corresponding increase in output entropy, suggesting the model remains confident even when it fails to use mid-context cues.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16134
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Beyond Early-Token Bias: Model-Specific and Language-Specific Position Effects in Multilingual LLMs
Menschikov, Mikhail
Kharitonov, Alexander
Kotyga, Maiia
Porvatov, Vadim
Zhukovskaya, Anna
Kagramanyan, David
Shvetsov, Egor
Burnaev, Evgeny
Computation and Language
Machine Learning
Large Language Models (LLMs) exhibit position bias systematically underweighting information based on its location in the context but how this bias varies across languages and models remains unclear. We conduct a multilingual study across five typologically diverse languages (English, Russian, German, Hindi, Vietnamese) and five model architectures, analyzing how position bias interacts with prompting strategies and affects output entropy. Our key findings are: (1) Position bias is primarily model-driven but shows language-specific nuances. Notably, Qwen2.5-7B-Instruct, DeepSeek 7B Chat and Mistral 7B consistently favor late positions challenging the common assumption of universal early-token preference. (2) Explicitly instructing the model, in the presence of irrelevant distractors, that "the most relevant context to the query is marked as 1" unexpectedly reduces accuracy across all languages, questioning standard prompt-engineering practices. (3) Accuracy consistently drops most when relevant information appears in the middle of the context, yet this is not reflected in a corresponding increase in output entropy, suggesting the model remains confident even when it fails to use mid-context cues.
title Beyond Early-Token Bias: Model-Specific and Language-Specific Position Effects in Multilingual LLMs
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2505.16134