Reproduce “VLM-as-a-Judge for Person Re-Identification” 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

Bessie Dominguez-Dager, Miguel Cazorla, Vincent Lepetit

The authors use two frozen person-reidentification specialists to rank gallery images, then ask a frozen vision-language model to reorder their combined shortlist when their top matches disagree. On CHIRLA’s 22 identities, the gated system reaches 64.35% average rank-1 at pool size 16, 12.9 points above SapiensID; gating saves compute but reduces mAP.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.