MLLM-CompBench
MLLM-CompBench copied to clipboard
[NeurIPS'25] MLLM-CompBench evaluates the comparative reasoning of MLLMs with 40K image pairs and questions across 8 dimensions of relative comparison: visual attribute, existence, state, emotion, tem...
Two datasets cannot be processed due to misssing files / non-existing dirs, as shown below: 1. cub_200_2011    2. soccernet (I have run "preprocessing/soccernet/download_soccernet.py" to download all the...
To make data access easier, we provide processed data and updated annotations on both our project website and GitHub page. https://compbench.github.io/ https://github.com/RaptorMai/MLLM-CompBench Please open an issue if you have any...