FineVision, the 17M-image open VLM dataset from 200+ sources, accepted to NeurIPS

andimarafioti · x · 2026-09-25

FineVision — the open-source VLM dataset curation the team released earlier — has been accepted to NeurIPS. The dataset aggregates 200+ sources into 17M unique images and 10B answer tokens, adding new capabilities like GUI navigation, pointing, and counting.

According to the authors, training on FineVision improves open-source VLMs by 20% across 10 benchmarks. First authors lusxvr and orrzohar will present in Sydney.

Original post →

More from Multimodal

Multimodal channel →