JaneDing
@janeding.bsky.social
Data Science junior @Umich
· Starting research in vision-language models and pragmatic generation · Exploring how AI communicates like humans
homepage: jingding-ai.github.io
· Starting research in vision-language models and pragmatic generation · Exploring how AI communicates like humans
homepage: jingding-ai.github.io
Reposted by JaneDing
Vision-Language Models are not yet pragmatically optimal.
We identify 3 key failures of pragmatic competence in referring expression generation with VLMs: (1) cannot uniquely refer to the referent, (2) include excessive or irrelevant information, and (3) misalign with human pragmatic preferences.
We identify 3 key failures of pragmatic competence in referring expression generation with VLMs: (1) cannot uniquely refer to the referent, (2) include excessive or irrelevant information, and (3) misalign with human pragmatic preferences.
April 23, 2025 at 5:55 PM
Vision-Language Models are not yet pragmatically optimal.
We identify 3 key failures of pragmatic competence in referring expression generation with VLMs: (1) cannot uniquely refer to the referent, (2) include excessive or irrelevant information, and (3) misalign with human pragmatic preferences.
We identify 3 key failures of pragmatic competence in referring expression generation with VLMs: (1) cannot uniquely refer to the referent, (2) include excessive or irrelevant information, and (3) misalign with human pragmatic preferences.