arXiv cs.CL· Athira Gopal, Patil Prathamesh Pravin, Ankush Kadu, Arpit Jain, Ashwanth Krishnan·· 13 小时前AI 评分39
同一实体，不同文字：诊断语言模型中的跨文字实体匹配失败
Same Entity, Different Script: Diagnosing Cross-Script Entity Matching Failures in Language Models
AI 导读
多语言模型对同一人名在不同文字下会给出不同身份判断：Qwen2.5-7B 在马拉地语提示下，拉丁文字的同一人准确率为 35.1%，梵文天城体降至 8.5%，而不同人准确率仍高于 98%。
正文 · 原文
Abstract:Multilingual language models can give different identity judgments when the same people's names are written in different scripts. We study this effect using 601 audited identity pairs with attested Devanagari and Latin forms, holding the question and relation fixed. Qwen2.5-7B's same-person accuracy falls from 35.1% in Latin to 8.5% in Devanagari under Marathi prompts, while different-person accuracy remains above 98%. Evaluation across five models reveals distinct response biases, and reduced tokenization fragmentation does not reliably close the gap. Replicating the design on 374 audited Malayalam pairs built from the same entities reproduces the effect in a second script and language family: same-person accuracy falls to 5.3%, about half the matched Devanagari rate, while the margin still ranks matching above non-matching names at 0.925 AUROC. Latin-trained probes recover label information from Devanagari states, exceeding a string-feature baseline under similarity-matched negatives. Scalar calibration recovers 87% of the raw model-probe balanced-accuracy gap, identifying decision criterion placement as a major contributor. Activation patching localises causal sensitivity to the second entity in Qwen's middle layers; one shared sensitivity direction transfers about half the full-state margin correction. A fact-independent intervention at this site raises held-out Marathi same-person accuracy from 6.5% to 33.5%, with 98.5% different-person accuracy. Together, these results distinguish script-dependent discrimination from criterion placement and identify a causal site for improving identity judgments.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.31407 [cs.CL] |
| (or arXiv:2610.31407v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.31407 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Athira Gopal [view email]
[v1]
Thu, 22 Oct 2026 15:09:46 UTC (337 KB)
来源：arXiv cs.CL · arxiv.org