English(EN) The Mirage of Calibrated Confidence: Trajectory-Independence of Verbalized Confidence in Vision-Language Models

新研究质疑VLM置信度校准，提出新的评估指标

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-17 04:00

一篇新发表在arXiv上的研究论文《校准置信度的幻象》（The Mirage of Calibrated Confidence）揭示，视觉语言模型（VLM）通常会报告其答案的高置信度，而这与其遵循的推理过程无关。研究发现，口头置信度在很大程度上独立于模型的内部轨迹，这意味着它不能准确反映其步骤的正确性。为解决此问题，研究人员提出了一种名为轨迹基础评分（Trajectory-Grounding Score, TGS）的新评估指标，以及一个名为TGS-Bench的基准测试套件，旨在通过比较模型在有无访问其推理路径时的置信度水平，来更好地评估VLM的可靠性。 AI

影响强调了VLM评估中的一个关键缺陷，可能导致更可靠的模型和对其局限性的更好理解。

排序理由发表在arXiv上的研究论文，提出了一种新的视觉语言模型评估指标。[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: 13 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 发表在arXiv上的研究论文，提出了一种新的视觉语言模型评估指标。[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: 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) · Jisoo Yang, Jaeho Han, Trung X. Pham, Junyeong Kim · 2026-09-17 04:00

校准置信度的海市蜃楼：视觉语言模型口头置信度的轨迹独立性

arXiv:2609.18453v1 Announce Type: new Abstract: A calibrated Vision-Language Model (VLM) can repeatedly self-correct, say "Wait, I should recheck," arrive at the wrong answer, and still report high confidence. We find that this occurs because verbalized confidence is largely traj…

本文如何被排名

报道来源 [1]

校准置信度的海市蜃楼：视觉语言模型口头置信度的轨迹独立性

相关实体

相关话题