Archive
Multimodal & Vision

The Word For A Visual Prompt

Archive
Hard
200pts0 solves
Yang et al. (2023) proposed improving GPT-4V's grounding by overlaying numbered marks on an image to let the model refer to objects by index. Name the technique (three-word hyphenated or acronym). Flag format: CONGRESS{name}. Example: CONGRESS{red-circle}.
Show hint
A noun for 'group' + a noun for 'label'.

Archive — no submissions accepted

This challenge is preserved for reference. Play live challenges at /challenges.