{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:H2RKB7JIH46DKCHKTBGRZRK363","short_pith_number":"pith:H2RKB7JI","schema_version":"1.0","canonical_sha256":"3ea2a0fd283f3c3508ea984d1cc55bf6e55dcd2d1f9d8f3e3e191fd6dc48a38d","source":{"kind":"arxiv","id":"2501.06521","version":1},"attestation_state":"computed","paper":{"title":"Fine-tuning Large Language Models for Improving Factuality in Legal Question Answering","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fei Wu, Kun Kuang, Leilei Gan, Wenyi Xiao, Yinghao Hu","submitted_at":"2025-01-11T12:08:15Z","abstract_excerpt":"Hallucination, or the generation of incorrect or fabricated information, remains a critical challenge in large language models (LLMs), particularly in high-stake domains such as legal question answering (QA). In order to mitigate the hallucination rate in legal QA, we first introduce a benchmark called LegalHalBench and three automatic metrics to evaluate the common hallucinations when LLMs answer legal questions. We then propose a hallucination mitigation method that integrates behavior cloning and a novel Hard Sample-aware Iterative Direct Preference Optimization (HIPO). We conduct extensive"},"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":"2501.06521","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-01-11T12:08:15Z","cross_cats_sorted":[],"title_canon_sha256":"d843c98769656f6d0b9857b6691ee8bd75645febc863b5d2780528166f9f27a1","abstract_canon_sha256":"56757f50a3eadbcd19b8598031cf0d4f7a4f05443a12de998301b421b59a6d53"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:00:04.671646Z","signature_b64":"bhzgtv4wcAN78z8rwLoRrKBwuKhl8EImdcMRIyYFzF9Mu0yn1ifLxKxu7KVwX+Q7vGwd4fMCZRK6t5VFGHsQAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3ea2a0fd283f3c3508ea984d1cc55bf6e55dcd2d1f9d8f3e3e191fd6dc48a38d","last_reissued_at":"2026-07-05T10:00:04.671099Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:00:04.671099Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fine-tuning Large Language Models for Improving Factuality in Legal Question Answering","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fei Wu, Kun Kuang, Leilei Gan, Wenyi Xiao, Yinghao Hu","submitted_at":"2025-01-11T12:08:15Z","abstract_excerpt":"Hallucination, or the generation of incorrect or fabricated information, remains a critical challenge in large language models (LLMs), particularly in high-stake domains such as legal question answering (QA). In order to mitigate the hallucination rate in legal QA, we first introduce a benchmark called LegalHalBench and three automatic metrics to evaluate the common hallucinations when LLMs answer legal questions. We then propose a hallucination mitigation method that integrates behavior cloning and a novel Hard Sample-aware Iterative Direct Preference Optimization (HIPO). We conduct extensive"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.06521","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/2501.06521/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":"2501.06521","created_at":"2026-07-05T10:00:04.671162+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.06521v1","created_at":"2026-07-05T10:00:04.671162+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.06521","created_at":"2026-07-05T10:00:04.671162+00:00"},{"alias_kind":"pith_short_12","alias_value":"H2RKB7JIH46D","created_at":"2026-07-05T10:00:04.671162+00:00"},{"alias_kind":"pith_short_16","alias_value":"H2RKB7JIH46DKCHK","created_at":"2026-07-05T10:00:04.671162+00:00"},{"alias_kind":"pith_short_8","alias_value":"H2RKB7JI","created_at":"2026-07-05T10:00:04.671162+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.06272","citing_title":"Evaluating LLM-based Approaches to Legal Citation Prediction: Domain-specific Pre-training, Fine-tuning, or RAG? A Benchmark and an Australian Law Case Study","ref_index":21,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/H2RKB7JIH46DKCHKTBGRZRK363","json":"https://pith.science/pith/H2RKB7JIH46DKCHKTBGRZRK363.json","graph_json":"https://pith.science/api/pith-number/H2RKB7JIH46DKCHKTBGRZRK363/graph.json","events_json":"https://pith.science/api/pith-number/H2RKB7JIH46DKCHKTBGRZRK363/events.json","paper":"https://pith.science/paper/H2RKB7JI"},"agent_actions":{"view_html":"https://pith.science/pith/H2RKB7JIH46DKCHKTBGRZRK363","download_json":"https://pith.science/pith/H2RKB7JIH46DKCHKTBGRZRK363.json","view_paper":"https://pith.science/paper/H2RKB7JI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.06521&json=true","fetch_graph":"https://pith.science/api/pith-number/H2RKB7JIH46DKCHKTBGRZRK363/graph.json","fetch_events":"https://pith.science/api/pith-number/H2RKB7JIH46DKCHKTBGRZRK363/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/H2RKB7JIH46DKCHKTBGRZRK363/action/timestamp_anchor","attest_storage":"https://pith.science/pith/H2RKB7JIH46DKCHKTBGRZRK363/action/storage_attestation","attest_author":"https://pith.science/pith/H2RKB7JIH46DKCHKTBGRZRK363/action/author_attestation","sign_citation":"https://pith.science/pith/H2RKB7JIH46DKCHKTBGRZRK363/action/citation_signature","submit_replication":"https://pith.science/pith/H2RKB7JIH46DKCHKTBGRZRK363/action/replication_record"}},"created_at":"2026-07-05T10:00:04.671162+00:00","updated_at":"2026-07-05T10:00:04.671162+00:00"}