Reproduce “Bridging Modality Disconnect in Self-Reflection via Closed-Loop Visually Grounded Verification” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Haoyu Zhang, Yuwei Wu, Pengxiang Li, Xintong Zhang, Zhi Gao, Rui Gao, Mingyang Gao, Che Sun and 1 more
MIRROR helps vision-language models revisit uncertain answers by checking image regions before revising them. It identifies a suspected error, asks an external visual prompt generator to mark a relevant region, then uses the marked image and dialogue history to reconsider its response. On MM-Vet, the loop scores 66.70 versus 62.70 without it, but this comparison does not isolate all causes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.