{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:BDWJUVQWQAL75FUDDZPQYMJM64","short_pith_number":"pith:BDWJUVQW","schema_version":"1.0","canonical_sha256":"08ec9a56168017fe96831e5f0c312cf72a4c6805fd26b3f3d594baa7cf51404c","source":{"kind":"arxiv","id":"2311.09627","version":2},"attestation_state":"computed","paper":{"title":"Mitigating Biases for Instruction-following Language Models via Bias Neurons Elimination","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Honglak Lee, Jungkyu Choi, Kyomin Jung, Nakyeong Yang, Taegwan Kang","submitted_at":"2023-11-16T07:16:55Z","abstract_excerpt":"Instruction-following language models often show undesirable biases. These undesirable biases may be accelerated in the real-world usage of language models, where a wide range of instructions is used through zero-shot example prompting. To solve this problem, we first define the bias neuron, which significantly affects biased outputs, and prove its existence empirically. Furthermore, we propose a novel and practical bias mitigation method, CRISPR, to eliminate bias neurons of language models in instruction-following settings. CRISPR automatically determines biased outputs and categorizes neuro"},"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":"2311.09627","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2023-11-16T07:16:55Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"14b1fe38f66d0944a9665ae69e8effe574298c106b813bd0abac4cea53a4f929","abstract_canon_sha256":"db152791629a4582c22bd6ec14ef782855404a958abc11077b8e6eca846c3660"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:27:23.610614Z","signature_b64":"GG3fmKclUD23fZ9PAXsPRf+Xd9XwNg5Wwq1xMzsccZ+TMndXruTowOphMhh7LIcmFMOY5hzvYg6zl09aKRrVCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"08ec9a56168017fe96831e5f0c312cf72a4c6805fd26b3f3d594baa7cf51404c","last_reissued_at":"2026-07-05T08:27:23.610193Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:27:23.610193Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mitigating Biases for Instruction-following Language Models via Bias Neurons Elimination","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Honglak Lee, Jungkyu Choi, Kyomin Jung, Nakyeong Yang, Taegwan Kang","submitted_at":"2023-11-16T07:16:55Z","abstract_excerpt":"Instruction-following language models often show undesirable biases. These undesirable biases may be accelerated in the real-world usage of language models, where a wide range of instructions is used through zero-shot example prompting. To solve this problem, we first define the bias neuron, which significantly affects biased outputs, and prove its existence empirically. Furthermore, we propose a novel and practical bias mitigation method, CRISPR, to eliminate bias neurons of language models in instruction-following settings. CRISPR automatically determines biased outputs and categorizes neuro"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.09627","kind":"arxiv","version":2},"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/2311.09627/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":"2311.09627","created_at":"2026-07-05T08:27:23.610254+00:00"},{"alias_kind":"arxiv_version","alias_value":"2311.09627v2","created_at":"2026-07-05T08:27:23.610254+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.09627","created_at":"2026-07-05T08:27:23.610254+00:00"},{"alias_kind":"pith_short_12","alias_value":"BDWJUVQWQAL7","created_at":"2026-07-05T08:27:23.610254+00:00"},{"alias_kind":"pith_short_16","alias_value":"BDWJUVQWQAL75FUD","created_at":"2026-07-05T08:27:23.610254+00:00"},{"alias_kind":"pith_short_8","alias_value":"BDWJUVQW","created_at":"2026-07-05T08:27:23.610254+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/BDWJUVQWQAL75FUDDZPQYMJM64","json":"https://pith.science/pith/BDWJUVQWQAL75FUDDZPQYMJM64.json","graph_json":"https://pith.science/api/pith-number/BDWJUVQWQAL75FUDDZPQYMJM64/graph.json","events_json":"https://pith.science/api/pith-number/BDWJUVQWQAL75FUDDZPQYMJM64/events.json","paper":"https://pith.science/paper/BDWJUVQW"},"agent_actions":{"view_html":"https://pith.science/pith/BDWJUVQWQAL75FUDDZPQYMJM64","download_json":"https://pith.science/pith/BDWJUVQWQAL75FUDDZPQYMJM64.json","view_paper":"https://pith.science/paper/BDWJUVQW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2311.09627&json=true","fetch_graph":"https://pith.science/api/pith-number/BDWJUVQWQAL75FUDDZPQYMJM64/graph.json","fetch_events":"https://pith.science/api/pith-number/BDWJUVQWQAL75FUDDZPQYMJM64/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BDWJUVQWQAL75FUDDZPQYMJM64/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BDWJUVQWQAL75FUDDZPQYMJM64/action/storage_attestation","attest_author":"https://pith.science/pith/BDWJUVQWQAL75FUDDZPQYMJM64/action/author_attestation","sign_citation":"https://pith.science/pith/BDWJUVQWQAL75FUDDZPQYMJM64/action/citation_signature","submit_replication":"https://pith.science/pith/BDWJUVQWQAL75FUDDZPQYMJM64/action/replication_record"}},"created_at":"2026-07-05T08:27:23.610254+00:00","updated_at":"2026-07-05T08:27:23.610254+00:00"}