{"paper":{"title":"Shortcut Mitigation via Spurious-Positive Samples","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"Identifying a small set of instances where models rely on spurious attributes and regularizing the associated neurons improves robustness without needing extra annotations or balanced data.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Christin Seifert, Gemma Roig, J\\\"org Schl\\\"otterer, Phuong Quynh Le, Sari Sadiya","submitted_at":"2026-05-13T11:00:12Z","abstract_excerpt":"Shortcut mitigation strategies commonly rely on training data annotations, group-balanced held-out data or the presence of all groups, i.e., all combinations of (spurious) attributes and classes, in the training data. However, these requirements are rarely met in practice. We instead propose a method for targeted model analysis to identify a small set of instances in which the model relies on spurious attributes. Using that set and following ``this feature should not be used for prediction'' reasoning, we identify highly relevant neurons in an intermediate layer and regularize their impact. Th"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"This ensures that models learn to depend on informative features rather than being right for the wrong reasons, thereby improving robustness without requiring additional balanced held-out data or annotations.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That a small set of instances can be reliably identified where the model relies on spurious attributes, and that regularizing the corresponding neurons will sufficiently reduce shortcut dependence without harming performance on core features.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A method uses spurious-positive samples to identify and regularize neurons that rely on spurious features, improving model robustness without extra annotations or balanced data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Identifying a small set of instances where models rely on spurious attributes and regularizing the associated neurons improves robustness without needing extra annotations or balanced data.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ef4e6f7ab459ccae856298c67102e8ca43bbc81dff50984b7dbd0ee6827818ed"},"source":{"id":"2605.13340","kind":"arxiv","version":1},"verdict":{"id":"d2faf604-2f3d-46c7-8877-f0093ed612df","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:10:11.829793Z","strongest_claim":"This ensures that models learn to depend on informative features rather than being right for the wrong reasons, thereby improving robustness without requiring additional balanced held-out data or annotations.","one_line_summary":"A method uses spurious-positive samples to identify and regularize neurons that rely on spurious features, improving model robustness without extra annotations or balanced data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That a small set of instances can be reliably identified where the model relies on spurious attributes, and that regularizing the corresponding neurons will sufficiently reduce shortcut dependence without harming performance on core features.","pith_extraction_headline":"Identifying a small set of instances where models rely on spurious attributes and regularizing the associated neurons improves robustness without needing extra annotations or balanced data."},"references":{"count":36,"sample":[{"doi":"","year":2024,"title":"In: Salakhutdinov, R., Kolter, Z., Heller, K., Weller, A., Oliver, N., Scarlett, J., Berkenkamp, F","work_id":"33a3647b-4ff4-49c3-9ac5-f80f73c5a62e","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Advances in Neural Information Processing Systems35, 23284–23296 (2022)","work_id":"8b055e90-77e3-4904-a929-719fc8b5e02e","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1371/journal.pone.0130140","year":2015,"title":"On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer- Wise Relevance Propagation","work_id":"a2492a93-db2c-4bde-8872-6eaf0d7e310a","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Advances in Neural Information Processing Systems37, 106383–106410 (2024)","work_id":"963b5444-d8ba-4082-8b06-af33b5fb8d3a","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Advances in Neural Information Processing Systems35, 33618– 33632 (2022)","work_id":"8c8dea8c-d605-4bb6-b0ec-d98e8acb5176","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":36,"snapshot_sha256":"c34f9fb0e604f30139315003d47362e6944a6d50211d984299891862f1936065","internal_anchors":2},"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"}