A four-metric suite shows VLM self-play in referential games diverges from human grounding patterns, with GPT4o-mini closest and task success not implying common ground.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Measuring How (Not Just Whether) VLMs Build Common Ground
A four-metric suite shows VLM self-play in referential games diverges from human grounding patterns, with GPT4o-mini closest and task success not implying common ground.