Early-Signal Detector

phenomenon · first seen 16 Nov 2026

vision-language model visual attention planning

Also: VLM gaze planning

How vision-language models choose the order of image regions to describe, studied via internal attention mechanisms.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

activation patching