{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ZGZYN5KFZRC4DTYMACJ3XVYVV4","short_pith_number":"pith:ZGZYN5KF","schema_version":"1.0","canonical_sha256":"c9b386f545cc45c1cf0c0093bbd715af2844106460463f8d3a35ec07a1e3efd5","source":{"kind":"arxiv","id":"2505.23410","version":1},"attestation_state":"computed","paper":{"title":"From Parameters to Prompts: Understanding and Mitigating the Factuality Gap between Fine-Tuned LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hanbo Huang, Shiyu Liang, Xuan Gong","submitted_at":"2025-05-29T12:59:30Z","abstract_excerpt":"Factual knowledge extraction aims to explicitly extract knowledge parameterized in pre-trained language models for application in downstream tasks. While prior work has been investigating the impact of supervised fine-tuning data on the factuality of large language models (LLMs), its mechanism remains poorly understood. We revisit this impact through systematic experiments, with a particular focus on the factuality gap that arises when fine-tuning on known versus unknown knowledge. Our findings show that this gap can be mitigated at the inference stage, either under out-of-distribution (OOD) s"},"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":"2505.23410","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-29T12:59:30Z","cross_cats_sorted":[],"title_canon_sha256":"9e8ccf9a863e6fe9d0a7a565f1672d09c70d50da9382d1a7aff2cde42155e470","abstract_canon_sha256":"1f428ca72891d9aeee92fa4988a93b6859ecb864a1af8c6d56236cbcbe33b774"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:12:01.268822Z","signature_b64":"rcttB71esEUSZ7c/vF5oLBtn+ACUzwEq7fZC2NOkrme+k2yHNFbH/ulRkQCKMBvZgaCQCLCDEAttI5jaFiiICQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c9b386f545cc45c1cf0c0093bbd715af2844106460463f8d3a35ec07a1e3efd5","last_reissued_at":"2026-07-05T11:12:01.268335Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:12:01.268335Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"From Parameters to Prompts: Understanding and Mitigating the Factuality Gap between Fine-Tuned LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hanbo Huang, Shiyu Liang, Xuan Gong","submitted_at":"2025-05-29T12:59:30Z","abstract_excerpt":"Factual knowledge extraction aims to explicitly extract knowledge parameterized in pre-trained language models for application in downstream tasks. While prior work has been investigating the impact of supervised fine-tuning data on the factuality of large language models (LLMs), its mechanism remains poorly understood. We revisit this impact through systematic experiments, with a particular focus on the factuality gap that arises when fine-tuning on known versus unknown knowledge. Our findings show that this gap can be mitigated at the inference stage, either under out-of-distribution (OOD) s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.23410","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/2505.23410/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":"2505.23410","created_at":"2026-07-05T11:12:01.268389+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.23410v1","created_at":"2026-07-05T11:12:01.268389+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.23410","created_at":"2026-07-05T11:12:01.268389+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZGZYN5KFZRC4","created_at":"2026-07-05T11:12:01.268389+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZGZYN5KFZRC4DTYM","created_at":"2026-07-05T11:12:01.268389+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZGZYN5KF","created_at":"2026-07-05T11:12:01.268389+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/ZGZYN5KFZRC4DTYMACJ3XVYVV4","json":"https://pith.science/pith/ZGZYN5KFZRC4DTYMACJ3XVYVV4.json","graph_json":"https://pith.science/api/pith-number/ZGZYN5KFZRC4DTYMACJ3XVYVV4/graph.json","events_json":"https://pith.science/api/pith-number/ZGZYN5KFZRC4DTYMACJ3XVYVV4/events.json","paper":"https://pith.science/paper/ZGZYN5KF"},"agent_actions":{"view_html":"https://pith.science/pith/ZGZYN5KFZRC4DTYMACJ3XVYVV4","download_json":"https://pith.science/pith/ZGZYN5KFZRC4DTYMACJ3XVYVV4.json","view_paper":"https://pith.science/paper/ZGZYN5KF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.23410&json=true","fetch_graph":"https://pith.science/api/pith-number/ZGZYN5KFZRC4DTYMACJ3XVYVV4/graph.json","fetch_events":"https://pith.science/api/pith-number/ZGZYN5KFZRC4DTYMACJ3XVYVV4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZGZYN5KFZRC4DTYMACJ3XVYVV4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZGZYN5KFZRC4DTYMACJ3XVYVV4/action/storage_attestation","attest_author":"https://pith.science/pith/ZGZYN5KFZRC4DTYMACJ3XVYVV4/action/author_attestation","sign_citation":"https://pith.science/pith/ZGZYN5KFZRC4DTYMACJ3XVYVV4/action/citation_signature","submit_replication":"https://pith.science/pith/ZGZYN5KFZRC4DTYMACJ3XVYVV4/action/replication_record"}},"created_at":"2026-07-05T11:12:01.268389+00:00","updated_at":"2026-07-05T11:12:01.268389+00:00"}