{"paper":{"title":"Wasserstein-Aligned Localisation for VLM-Based Distributional OOD Detection in Medical Imaging","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Wasserstein distances let vision-language models pick moderately similar normal anatomy references to localize anomalies in medical images.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bernhard Kainz, Cosmin Bercea, Johanna P Mueller, Matthew Baugh","submitted_at":"2026-05-06T17:32:34Z","abstract_excerpt":"Zero-shot anomaly localisation via vision-language models (VLMs) offers a compelling approach for rare pathology detection, yet its performance is fundamentally limited by the absence of healthy anatomical context. We reformulate zero-shot localisation as a comparative inference problem in which anomalies are identified through structured comparison against reference distributions of normal anatomy. We introduce WALDO, a training-free framework grounded in optimal transport theory that enables comparative reasoning through: (i) entropy-weighted Sliced Wasserstein distances for anatomically-awa"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On the NOVA brain MRI benchmark, WALDO with Qwen2.5-VL-72B achieves 43.5±1.6% mAP@30 (95% CI: [40.4, 46.7]), representing a 19% relative improvement over zero-shot baselines. Cross-model evaluation shows consistent gains: GPT-4o achieves 32.0±6.5% and Qwen3-VL-32B achieves 32.0±6.6% mAP@30. Paired McNemar tests confirm statistical significance (p<0.01).","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That reference images drawn from DINOv2 patch distributions of normal anatomy, selected via moderate similarity in the Goldilocks zone, provide a reliable bias-variance trade-off for comparative visual reasoning in VLMs, and that the non-monotonic relationship between reference similarity and localisation accuracy holds across the tested models and data.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"WALDO improves zero-shot anomaly localization in medical imaging by selecting reference distributions via entropy-weighted Sliced Wasserstein distances and Goldilocks zone sampling, yielding a 19% relative gain on brain MRI benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Wasserstein distances let vision-language models pick moderately similar normal anatomy references to localize anomalies in medical images.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e902904c901851938bdce7d34a7255cd501cc22ee62d5fcc2db09ae776337df1"},"source":{"id":"2605.05161","kind":"arxiv","version":2},"verdict":{"id":"d0712e97-fdef-4779-b88f-7808776b9384","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T16:36:15.678369Z","strongest_claim":"On the NOVA brain MRI benchmark, WALDO with Qwen2.5-VL-72B achieves 43.5±1.6% mAP@30 (95% CI: [40.4, 46.7]), representing a 19% relative improvement over zero-shot baselines. Cross-model evaluation shows consistent gains: GPT-4o achieves 32.0±6.5% and Qwen3-VL-32B achieves 32.0±6.6% mAP@30. Paired McNemar tests confirm statistical significance (p<0.01).","one_line_summary":"WALDO improves zero-shot anomaly localization in medical imaging by selecting reference distributions via entropy-weighted Sliced Wasserstein distances and Goldilocks zone sampling, yielding a 19% relative gain on brain MRI benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That reference images drawn from DINOv2 patch distributions of normal anatomy, selected via moderate similarity in the Goldilocks zone, provide a reliable bias-variance trade-off for comparative visual reasoning in VLMs, and that the non-monotonic relationship between reference similarity and localisation accuracy holds across the tested models and data.","pith_extraction_headline":"Wasserstein distances let vision-language models pick moderately similar normal anatomy references to localize anomalies in medical images."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.05161/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T10:35:54.023754Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T21:01:19.796836Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T13:47:58.196713Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"4e202d90f21440286011b3652365686311dd32a0de03ae4f5211e11989f8749d"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}