{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:YADLB7OIYQMOMM45WPRHSVO6SK","short_pith_number":"pith:YADLB7OI","schema_version":"1.0","canonical_sha256":"c006b0fdc8c418e6339db3e27955de92b4a853ba3153017aa64009d21fb71542","source":{"kind":"arxiv","id":"2410.22517","version":1},"attestation_state":"computed","paper":{"title":"Attention Speaks Volumes: Localizing and Mitigating Bias in Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Besmira Nushi, Rishabh Adiga, Varun Chandrasekaran","submitted_at":"2024-10-29T20:15:56Z","abstract_excerpt":"We explore the internal mechanisms of how bias emerges in large language models (LLMs) when provided with ambiguous comparative prompts: inputs that compare or enforce choosing between two or more entities without providing clear context for preference. Most approaches for bias mitigation focus on either post-hoc analysis or data augmentation. However, these are transient solutions, without addressing the root cause: the model itself. Numerous prior works show the influence of the attention module towards steering generations. We believe that analyzing attention is also crucial for understandi"},"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":"2410.22517","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-29T20:15:56Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"57e3b7ec7919d8900fb73bdc46711eb0020f6e8018e76ff8bb1351cf7cb6eb69","abstract_canon_sha256":"c077688005747ec76a4c99a6f770b657ed0cd0906a4e65a5350d953cb3d1d89d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:28:17.623772Z","signature_b64":"9XIo7iso/1mV/yTr4HanpqR6XaOHNX15QOomLorRhQ3EKbugDupe50S/k8DCZWitmuMXeyjbX/mS7LbxAPpCAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c006b0fdc8c418e6339db3e27955de92b4a853ba3153017aa64009d21fb71542","last_reissued_at":"2026-07-05T09:28:17.623330Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:28:17.623330Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Attention Speaks Volumes: Localizing and Mitigating Bias in Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Besmira Nushi, Rishabh Adiga, Varun Chandrasekaran","submitted_at":"2024-10-29T20:15:56Z","abstract_excerpt":"We explore the internal mechanisms of how bias emerges in large language models (LLMs) when provided with ambiguous comparative prompts: inputs that compare or enforce choosing between two or more entities without providing clear context for preference. Most approaches for bias mitigation focus on either post-hoc analysis or data augmentation. However, these are transient solutions, without addressing the root cause: the model itself. Numerous prior works show the influence of the attention module towards steering generations. We believe that analyzing attention is also crucial for understandi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.22517","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/2410.22517/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":"2410.22517","created_at":"2026-07-05T09:28:17.623389+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.22517v1","created_at":"2026-07-05T09:28:17.623389+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.22517","created_at":"2026-07-05T09:28:17.623389+00:00"},{"alias_kind":"pith_short_12","alias_value":"YADLB7OIYQMO","created_at":"2026-07-05T09:28:17.623389+00:00"},{"alias_kind":"pith_short_16","alias_value":"YADLB7OIYQMOMM45","created_at":"2026-07-05T09:28:17.623389+00:00"},{"alias_kind":"pith_short_8","alias_value":"YADLB7OI","created_at":"2026-07-05T09:28:17.623389+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.07771","citing_title":"Breaking Down Bias: On The Limits of Generalizable Pruning Strategies","ref_index":1,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YADLB7OIYQMOMM45WPRHSVO6SK","json":"https://pith.science/pith/YADLB7OIYQMOMM45WPRHSVO6SK.json","graph_json":"https://pith.science/api/pith-number/YADLB7OIYQMOMM45WPRHSVO6SK/graph.json","events_json":"https://pith.science/api/pith-number/YADLB7OIYQMOMM45WPRHSVO6SK/events.json","paper":"https://pith.science/paper/YADLB7OI"},"agent_actions":{"view_html":"https://pith.science/pith/YADLB7OIYQMOMM45WPRHSVO6SK","download_json":"https://pith.science/pith/YADLB7OIYQMOMM45WPRHSVO6SK.json","view_paper":"https://pith.science/paper/YADLB7OI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.22517&json=true","fetch_graph":"https://pith.science/api/pith-number/YADLB7OIYQMOMM45WPRHSVO6SK/graph.json","fetch_events":"https://pith.science/api/pith-number/YADLB7OIYQMOMM45WPRHSVO6SK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YADLB7OIYQMOMM45WPRHSVO6SK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YADLB7OIYQMOMM45WPRHSVO6SK/action/storage_attestation","attest_author":"https://pith.science/pith/YADLB7OIYQMOMM45WPRHSVO6SK/action/author_attestation","sign_citation":"https://pith.science/pith/YADLB7OIYQMOMM45WPRHSVO6SK/action/citation_signature","submit_replication":"https://pith.science/pith/YADLB7OIYQMOMM45WPRHSVO6SK/action/replication_record"}},"created_at":"2026-07-05T09:28:17.623389+00:00","updated_at":"2026-07-05T09:28:17.623389+00:00"}