English(EN) When Prompts Ignore Structure: Graph-Based Attribute Reasoning for Calibrated VLMs

新的ARGTCA方法提高了视觉语言模型的可靠性

作者 PulseAugur 编辑部 · [2 个来源] · 2026-07-08 13:31

研究人员开发了ARGTCA，一种用于提高视觉语言模型（VLMs）可靠性和置信度估计的新方法。该方法利用图注意力网络（GAT）来模拟类别属性之间的关系，解决了先前将属性视为独立的局限性。通过捕获属性间的依赖关系，ARGTCA旨在提高VLMs的校准性并减少其过度自信，尤其是在测试时适应期间。 AI

影响这项研究提供了一种新技术，通过改进其置信度估计来增强视觉语言模型的可靠性。

排序理由该集群包含一篇详细介绍改进AI模型新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

报道来源 [2]

arXiv cs.AI TIER_1 English(EN) · Tanay Sodha, Aditya Sharma, Ramya Hebbalaguppe, Vinti Agarwal, Pranav Murthy Yeluripaty · 2026-07-09 04:00

When Prompts Ignore Structure: Graph-Based Attribute Reasoning for Calibrated VLMs

arXiv:2607.07395v1 Announce Type: cross Abstract: Reliable confidence estimation remains a key limitation of test-time adaptation in vision-language models (VLMs), where prompt tuning improves zero-shot accuracy but often degrades calibration due to entropy-driven overconfidence.…
arXiv cs.AI TIER_1 English(EN) · Pranav Murthy Yeluripaty · 2026-07-08 13:31

When Prompts Ignore Structure: Graph-Based Attribute Reasoning for Calibrated VLMs

Reliable confidence estimation remains a key limitation of test-time adaptation in vision-language models (VLMs), where prompt tuning improves zero-shot accuracy but often degrades calibration due to entropy-driven overconfidence. Prior approaches mitigate this using LLM-derived …