VLM-as-a-judge
PulseAugur coverage of VLM-as-a-judge — every cluster mentioning VLM-as-a-judge across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
New PRISM framework uses category theory to refine AI-generated analogies
Researchers have introduced PRISM, a new framework grounded in category theory designed to measure and refine multimodal analogies. This system, evaluated on visual metaphor generation, uses a "pullback score" to quanti…
-
New benchmark AnnoBench evaluates visualization annotation generation
Researchers have introduced AnnoBench, a new benchmark designed to evaluate the automation of visualization annotation tasks. This benchmark addresses the challenge of generating annotations that meet visual, semantic, …
-
H-Adapter improves pose-robust hairstyle transfer using attention-derived masks
Researchers have developed H-Adapter, a novel system designed to improve the accuracy and robustness of hairstyle transfer, particularly when source and reference images have significant differences in head pose. The sy…