arXiv：cs.CL· Yun-Ang Wu, Xanh Ho, Andre Greiner-Petter, Sunisth Kumar, Tian Cheng Xia, Florian Boudin, Akiko Aizawa·· 4 小时前AI 评分36
多模态声明验证对 LLM 改写有多鲁棒？
How Robust Is Multimodal Claim Verification to LLM Rewriting?
AI 导读
研究测试了 11 个开源权重模型（覆盖 5 个 VLM 家族、2B 至 38B 参数）在多模态声明验证任务中对 LLM 改写的鲁棒性。结果显示多数模型准确率无显著下降，但概率偏移持续存在：对冲类改写几乎在所有模型上引发显著偏移，增强类效果较弱，语法纠错等常规润色影响很小。该工作已被 AACL 2026 主会接收。
正文
Abstract:LLMs are known to introduce stylistic changes into generated text, yet how these stylistic shifts affect model decisions on scientific tasks remains underexplored. In this paper, we focus on multimodal claim verification, where the goal is to determine whether a textual claim is grounded in a given piece of evidence. We apply two rewriting strategies: natural rewriting, which simulates how researchers routinely use LLMs to polish academic text, and controlled injection, which inserts a single LLM-associated word to isolate the effect of vocabulary choice. We evaluate 11 open-weight models spanning five VLM families and ranging from 2B to 38B parameters. We find that models are robust to these modifications: most show no significant drop in accuracy, and compared to prior work on review-score manipulation, verification appears far more stable. However, consistent probability shifts do occur. Hedging-oriented conditions produce significant shifts across nearly all models, while boosting conditions show a weaker effect and general polishing conditions (e.g., grammar correction, fluency improvement) have little effect.
| Comments: | Accepted to AACL 2026 (Main Conference). 18 pages |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.02841 [cs.CL] |
| (or arXiv:2610.02841v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.02841 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yun-Ang Wu [view email]
[v1]
Fri, 2 Oct 2026 05:31:40 UTC (1,399 KB)
来源：arXiv：cs.CL · arxiv.org