arXiv:2609.37863v1 Announce Type: cross
Abstract: Vision-language models (VLMs) are increasingly used in place of human annotators, making it important that substitutability tests reflect the model r...
By Nagham Omar, Mahmoud Jabarin, Kinan Ibraheem, Lotem Peled-Cohen
The paper introduces an audit layer for vision‑language model benchmarks that reuses per‑question artifacts from evaluation runs to assess the benchmark itself. Six detectors analyze saved outputs, and an additional text‑only rerun removes visual inputs to produce per‑question reports and aggregate statistics, including permutation‑null excess and interval estimates. Applying this method to the MMStar benchmark reveals that 20.2% of its questions show no accuracy gain from images, with a small but significant excess in mathematics and science categories, and identifies questions where all models chose the same non‑gold option.
By Denis Degtiarev, Anna Gromova, Sergey Katkov, Artem Mikhailin, Safina Mukhamedshina, Albert Nasybullin, Karina Shaikhutdinova, Artem Tuzov
VisionQ is a new benchmark for qualitative analysis in computer vision that evaluates vision‑language models (VLMs) on criterion‑conditioned visual discrimination. It is built from over 1,800 peer‑reviewed comparison figures in CVPR and ICCV papers, linking each image crop to author‑stated visual claims through 3,911 hand‑annotated data points. The benchmark includes a 51‑leaf taxonomy of visual criteria, a protocol that hides method identities and reports accuracy per criterion, and a DPO‑tuned Gemma‑4‑E4B judge that improves accuracy on a held‑out test set.
By Vu Dinh Xuan, Duc-Hai Nguyen, Minh-Dung Dao, Vu Quynh Giao, Quang Hong Nguyen, Binh-Son Hua, Barry O'Sullivan, David Murphy, Hoang D. Nguyen
The paper introduces a dual‑judge evaluation protocol for vision‑language models in legally grounded tasks, pairing a 0‑10 quality judge with a strict binary semantic‑equivalence judge. Using a controlled UK traffic‑sign interpretation task, the authors analyze 4,680 evaluations across visibility and occlusion conditions, finding moderate association between judges and an asymmetric Type II error pattern that is most pronounced under heavy occlusion. The protocol requires only one additional LLM call and reveals quality‑trustworthiness signals that single‑judge methods miss.
By Su Myat Noe, Ha Thanh Nguyen, May Myo Zin, Ken Satoh
JudgeSense is a benchmark comprising 880 items from human‑labelled corpora, each presented under two differently worded instructions that ask the same question. The study evaluates 25 judges from six providers across four tasks, measuring how rewording affects agreement with the judge’s own verdicts. Results show that rewording reduces agreement on all tasks, with significant effects on two, and that stability varies across tasks and is not predicted by parameter count.
By Rohith Reddy Bellibatlu, Edward Raff, Wenbin Zhang
JudgeProfile is a framework that analyzes the subjectivity of large language model (LLM) judges by separating evaluation into perception—how judges compare responses on attributes such as clarity, correctness, and detail—and prioritization—how much each attribute influences the final decision. Using the curated SubjectiveSet dataset of 50,013 response pairs evaluated by 21 judges across 87 attributes, the study finds that judges often agree on attribute judgments even when their overall choices differ. By estimating and adjusting attribute weights, the authors improve agreement with reference labels from 66.48% to 71.97%, outperforming fine‑tuning and rubric prompting.
By Qi Cao, Kangning Liu, Xuan Kan, Shunwen Tan, Yang Pei, Dake Chen, Yatai Ji, Zixuan Ye, Yuanpeng Tu, Daniel Li, Junbiao Tang, Pengtao Xie, Zihao He