Reproduce “CounterFactual Retrieval for Evidence-Aware Individual Animal 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
Ankit K. Upadhyay, Aleksandr Algasov, Ekaterina Nepovinnykh, Fedor Zolotarev, Tuomas Eerola, Heikki Kälviäinen, Charles V. Stewart, Lukas Picek
The authors introduce Counterfactual Foreground Retrieval (CFR), an evaluation method that uses masks to compare animal retrieval on original, foreground-only, and shortcut-only images while keeping the task fixed. It reports foreground support and shortcut recoverability rather than a single score. On humpback whales, two systems have similar full-image retrieval but markedly different shortcut recoverability; this diagnoses pixel-removal interventions, not model reasoning or causal sources.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.