English(EN) FailBench: How Reliable are VLMs at Judging Robot Task Success?

新的FailBench基准揭示了视觉语言模型在机器人任务成功评估方面的不可靠性

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-04 04:00

一个名为FailBench的新基准已被开发出来，用于评估视觉语言模型（VLMs）在判断机器人任务成功方面的可靠性。该基准包含2,197次操作尝试，结果显示，即使是表现最好的VLM，在检测机器人故障方面的平衡准确率也仅为0.77。针对故障检测进行微调的模型表现不如通用VLMs，并且在需要精细视觉证据的任务（如组装）上，它们的性能显著下降。研究人员发现，在模糊情况下存在预测成功的偏见，尽管空间定位等输入级干预显示出改进的潜力。 AI

影响突出了当前VLMs在关键机器人应用中的局限性，表明需要提高鲁棒性和进行专门的故障检测训练。

排序理由该集群包含一篇介绍新基准和评估结果的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 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: 17 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇介绍新基准和评估结果的学术论文。[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, product; 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: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Zaruhi Navasardyan, Tatul Danielyan, Hrant Davtyan · 2026-09-04 04:00

FailBench：视觉语言模型判断机器人任务成功率有多可靠？

arXiv:2609.03611v1 Announce Type: cross Abstract: Vision-Language Models (VLMs) are increasingly used to evaluate robot manipulation outcomes, but existing benchmarks offer limited evidence of cross-domain generalization. We introduce FailBench, a benchmark for robot failure dete…

本文如何被排名

报道来源 [1]

FailBench：视觉语言模型判断机器人任务成功率有多可靠？

相关实体

相关话题