Research
Zooming without Zooming v2 trains VLMs to connect visual evidence
EvidenceConnect synthesizes single-image questions requiring comparison, aggregation or chaining of zoomed regions, with independent VLM agreement and region-ablation filtering. ZwZ2 trained with RLVR reportedly improves fine-grained, spatial and long-context multimodal perception.
Every outlet that ran it 1 article
Hugging Face PapersResearchLead