English(EN) Do MLLM Judges Judge the Edit? Auditing Bias in Image Editing Evaluation with Verified Quality Preservation

新基准揭示多模态大语言模型在图像编辑评判中的偏见

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-02 04:00

研究人员开发了一个名为 EditJudgeBias 的新基准，用于审计多模态大语言模型（MLLM）在用作图像编辑任务裁判时的偏见。该基准包含 1,196 个真实编辑样本和 13 个旨在测试质量保留干预措施的注入线索。实验显示，MLLM 裁判容易受到无关视觉线索的影响，捏造的多数意见会提高评分，而候选顺序的改变会逆转相当一部分的决策。研究强调，需要不同的指标来表征 MLLM 裁判的鲁棒性，因为单一指标无法捕捉其性能的所有方面。 AI

影响凸显了基于大语言模型的评估系统中潜在的偏见，这对于可靠的 AI 训练和评估至关重要。

排序理由该集群包含一篇详细介绍新基准和实验结果的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CV 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 1 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍新基准和实验结果的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 1 days old; Coverage has settled into its steady-state source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.CV TIER_1 English(EN) · Yuan Huang, Zirui Song, Xiuying Chen · 2026-10-02 04:00

多模态大模型裁判是否会裁判编辑？通过验证的质量保留来审计图像编辑评估中的偏见

arXiv:2610.01670v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) are increasingly used as automated judges for instruction-based image editing and as reward signals for model training. However, systematically auditing whether these judges are influenced by…

本文如何被排名

报道来源 [1]

多模态大模型裁判是否会裁判编辑？通过验证的质量保留来审计图像编辑评估中的偏见

相关实体

相关话题