{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KNBKLFNN26IRXZZJBIWBNSAUQY","short_pith_number":"pith:KNBKLFNN","schema_version":"1.0","canonical_sha256":"5342a595add7911be7290a2c16c814863ca02c9c531fd77dd026fdb1b623115a","source":{"kind":"arxiv","id":"2505.18545","version":1},"attestation_state":"computed","paper":{"title":"B-score: Detecting biases in large language models using response history","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Anh Totti Nguyen, An Vo, Daeyoung Kim, Mohammad Reza Taesiri","submitted_at":"2025-05-24T06:23:52Z","abstract_excerpt":"Large language models (LLMs) often exhibit strong biases, e.g, against women or in favor of the number 7. We investigate whether LLMs would be able to output less biased answers when allowed to observe their prior answers to the same question in a multi-turn conversation. To understand which types of questions invite more biased answers, we test LLMs on our proposed set of questions that span 9 topics and belong to three types: (1) Subjective; (2) Random; and (3) Objective. Interestingly, LLMs are able to \"de-bias\" themselves in a multi-turn conversation in response to questions that seek an R"},"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":"2505.18545","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-24T06:23:52Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"714f6d568a7f66f653b8ab3571656e1dede6a00ce9cc785193595cdd0a8937c2","abstract_canon_sha256":"1644846fe159a238d585c2b1e72a9cd88bf768b881bf28d6bfdf7dc36e84918c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:08:48.176959Z","signature_b64":"r7k77+Q2B1Z1ICXpej0wZwBIKPlZj3m02/5C3vJO1vypNcVafh14rmizSfveZqrCmINY7dGewG+GfkaEys24BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5342a595add7911be7290a2c16c814863ca02c9c531fd77dd026fdb1b623115a","last_reissued_at":"2026-07-05T11:08:48.176476Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:08:48.176476Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"B-score: Detecting biases in large language models using response history","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Anh Totti Nguyen, An Vo, Daeyoung Kim, Mohammad Reza Taesiri","submitted_at":"2025-05-24T06:23:52Z","abstract_excerpt":"Large language models (LLMs) often exhibit strong biases, e.g, against women or in favor of the number 7. We investigate whether LLMs would be able to output less biased answers when allowed to observe their prior answers to the same question in a multi-turn conversation. To understand which types of questions invite more biased answers, we test LLMs on our proposed set of questions that span 9 topics and belong to three types: (1) Subjective; (2) Random; and (3) Objective. Interestingly, LLMs are able to \"de-bias\" themselves in a multi-turn conversation in response to questions that seek an R"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.18545","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/2505.18545/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":"2505.18545","created_at":"2026-07-05T11:08:48.176533+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.18545v1","created_at":"2026-07-05T11:08:48.176533+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.18545","created_at":"2026-07-05T11:08:48.176533+00:00"},{"alias_kind":"pith_short_12","alias_value":"KNBKLFNN26IR","created_at":"2026-07-05T11:08:48.176533+00:00"},{"alias_kind":"pith_short_16","alias_value":"KNBKLFNN26IRXZZJ","created_at":"2026-07-05T11:08:48.176533+00:00"},{"alias_kind":"pith_short_8","alias_value":"KNBKLFNN","created_at":"2026-07-05T11:08:48.176533+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.22714","citing_title":"AMEL: Accumulated Message Effects on LLM Judgments","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22714","citing_title":"AMEL: Accumulated Message Effects on LLM Judgments","ref_index":29,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KNBKLFNN26IRXZZJBIWBNSAUQY","json":"https://pith.science/pith/KNBKLFNN26IRXZZJBIWBNSAUQY.json","graph_json":"https://pith.science/api/pith-number/KNBKLFNN26IRXZZJBIWBNSAUQY/graph.json","events_json":"https://pith.science/api/pith-number/KNBKLFNN26IRXZZJBIWBNSAUQY/events.json","paper":"https://pith.science/paper/KNBKLFNN"},"agent_actions":{"view_html":"https://pith.science/pith/KNBKLFNN26IRXZZJBIWBNSAUQY","download_json":"https://pith.science/pith/KNBKLFNN26IRXZZJBIWBNSAUQY.json","view_paper":"https://pith.science/paper/KNBKLFNN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.18545&json=true","fetch_graph":"https://pith.science/api/pith-number/KNBKLFNN26IRXZZJBIWBNSAUQY/graph.json","fetch_events":"https://pith.science/api/pith-number/KNBKLFNN26IRXZZJBIWBNSAUQY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KNBKLFNN26IRXZZJBIWBNSAUQY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KNBKLFNN26IRXZZJBIWBNSAUQY/action/storage_attestation","attest_author":"https://pith.science/pith/KNBKLFNN26IRXZZJBIWBNSAUQY/action/author_attestation","sign_citation":"https://pith.science/pith/KNBKLFNN26IRXZZJBIWBNSAUQY/action/citation_signature","submit_replication":"https://pith.science/pith/KNBKLFNN26IRXZZJBIWBNSAUQY/action/replication_record"}},"created_at":"2026-07-05T11:08:48.176533+00:00","updated_at":"2026-07-05T11:08:48.176533+00:00"}