ECCV 2026 talk proposes desiderata for how LLMs should communicate uncertainty

abursuc · x · 2026-09-08

At an ECCV 2026 workshop (#uncv2026), Christian Hardmeier presented 'Towards Anthropomimetic Uncertainty: Desiderata for Uncertainty Communication in Large Language Models,' arguing that systems must express uncertainty in ways users can correctly interpret — uncertainty communication as an epistemic marker, not just internal quantification. A notable topic at the intersection of HCI and model trustworthiness.

Related event: ECCV 2026 workshop covers LLM uncertainty and multimodal diagnosis(3 posts)→

Original post →

More from Research

Research channel →