{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:JDKLNBOOYIXUCW5P63PVYMKT5B","short_pith_number":"pith:JDKLNBOO","schema_version":"1.0","canonical_sha256":"48d4b685cec22f415baff6df5c3153e850c30cde78e08a56de4742d83e68822d","source":{"kind":"arxiv","id":"2607.12863","version":1},"attestation_state":"computed","paper":{"title":"Toward Localizing and Repairing Bias in Transformer Attention Heads","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.SE","authors_text":"Sigma Jahan","submitted_at":"2026-07-14T15:15:16Z","abstract_excerpt":"Transformer language models are increasingly used as software components, yet biased outputs remain difficult to localize and repair inside the model. Existing fairness testing and repair methods largely operate at the input-output or retraining level, while recent work suggests that bias-related behavior can concentrate in a small set of attention heads. This paper studies whether attention heads can be localized and repaired through a targeted inference-time intervention. We introduce ROBIN, a white-box head-level fairness debugging method that ranks attention heads using sensitivity to fair"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2607.12863","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-07-14T15:15:16Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"f1cc6776de28a8083c366ff6aee56924c03125a953e5d2b85329c71099e6b7b3","abstract_canon_sha256":"f633172249fd650207050ce7c15acb0d95d7714ec3280935ba2779389f76f2ee"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-15T01:22:22.321965Z","signature_b64":"N7UIDe8qefzUdy8HMu8OpjFDNiUTHP5OZfCp2X/T5aLcpEVURFlPbApLgBfnWdSjC4tAFoiDQXD/aJP05Wn8DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"48d4b685cec22f415baff6df5c3153e850c30cde78e08a56de4742d83e68822d","last_reissued_at":"2026-07-15T01:22:22.321103Z","signature_status":"signed_v1","first_computed_at":"2026-07-15T01:22:22.321103Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Toward Localizing and Repairing Bias in Transformer Attention Heads","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.SE","authors_text":"Sigma Jahan","submitted_at":"2026-07-14T15:15:16Z","abstract_excerpt":"Transformer language models are increasingly used as software components, yet biased outputs remain difficult to localize and repair inside the model. Existing fairness testing and repair methods largely operate at the input-output or retraining level, while recent work suggests that bias-related behavior can concentrate in a small set of attention heads. This paper studies whether attention heads can be localized and repaired through a targeted inference-time intervention. We introduce ROBIN, a white-box head-level fairness debugging method that ranks attention heads using sensitivity to fair"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.12863","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2607.12863/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2607.12863","created_at":"2026-07-15T01:22:22.321554+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.12863v1","created_at":"2026-07-15T01:22:22.321554+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.12863","created_at":"2026-07-15T01:22:22.321554+00:00"},{"alias_kind":"pith_short_12","alias_value":"JDKLNBOOYIXU","created_at":"2026-07-15T01:22:22.321554+00:00"},{"alias_kind":"pith_short_16","alias_value":"JDKLNBOOYIXUCW5P","created_at":"2026-07-15T01:22:22.321554+00:00"},{"alias_kind":"pith_short_8","alias_value":"JDKLNBOO","created_at":"2026-07-15T01:22:22.321554+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JDKLNBOOYIXUCW5P63PVYMKT5B","json":"https://pith.science/pith/JDKLNBOOYIXUCW5P63PVYMKT5B.json","graph_json":"https://pith.science/api/pith-number/JDKLNBOOYIXUCW5P63PVYMKT5B/graph.json","events_json":"https://pith.science/api/pith-number/JDKLNBOOYIXUCW5P63PVYMKT5B/events.json","paper":"https://pith.science/paper/JDKLNBOO"},"agent_actions":{"view_html":"https://pith.science/pith/JDKLNBOOYIXUCW5P63PVYMKT5B","download_json":"https://pith.science/pith/JDKLNBOOYIXUCW5P63PVYMKT5B.json","view_paper":"https://pith.science/paper/JDKLNBOO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.12863&json=true","fetch_graph":"https://pith.science/api/pith-number/JDKLNBOOYIXUCW5P63PVYMKT5B/graph.json","fetch_events":"https://pith.science/api/pith-number/JDKLNBOOYIXUCW5P63PVYMKT5B/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JDKLNBOOYIXUCW5P63PVYMKT5B/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JDKLNBOOYIXUCW5P63PVYMKT5B/action/storage_attestation","attest_author":"https://pith.science/pith/JDKLNBOOYIXUCW5P63PVYMKT5B/action/author_attestation","sign_citation":"https://pith.science/pith/JDKLNBOOYIXUCW5P63PVYMKT5B/action/citation_signature","submit_replication":"https://pith.science/pith/JDKLNBOOYIXUCW5P63PVYMKT5B/action/replication_record"}},"created_at":"2026-07-15T01:22:22.321554+00:00","updated_at":"2026-07-15T01:22:22.321554+00:00"}