{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QQJX4ADBWGK7N2AVXVOLLMJGXG","short_pith_number":"pith:QQJX4ADB","schema_version":"1.0","canonical_sha256":"84137e0061b195f6e815bd5cb5b126b999dae2a611eacc03cc0060c79b349f3e","source":{"kind":"arxiv","id":"2405.05904","version":3},"attestation_state":"computed","paper":{"title":"Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amir Feder, Gal Yona, Jonathan Herzig, Matan Eyal, Roee Aharoni, Roi Reichart, Zorik Gekhman","submitted_at":"2024-05-09T17:00:22Z","abstract_excerpt":"When large language models are aligned via supervised fine-tuning, they may encounter new factual information that was not acquired through pre-training. It is often conjectured that this can teach the model the behavior of hallucinating factually incorrect responses, as the model is trained to generate facts that are not grounded in its pre-existing knowledge. In this work, we study the impact of such exposure to new knowledge on the capability of the fine-tuned model to utilize its pre-existing knowledge. To this end, we design a controlled setup, focused on closed-book QA, where we vary the"},"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":"2405.05904","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-09T17:00:22Z","cross_cats_sorted":[],"title_canon_sha256":"5cd90feb91ac6f1f80ea3050297136f10bf480fb3bf7509c77a54c4194c64edf","abstract_canon_sha256":"2c40d8604d48bd5325178f9f4cd5f409964dc51152e5719ed87c4279fbcb9f0a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:13:52.441999Z","signature_b64":"wjEdu4bQBSoaLAnOq7HUOylBrJIa+MkjFfKSgwhcbRwkYDjxIaHAPPN4qDijlYZWfmtQ2v8yYRYInywiZAIHAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"84137e0061b195f6e815bd5cb5b126b999dae2a611eacc03cc0060c79b349f3e","last_reissued_at":"2026-07-05T09:13:52.441554Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:13:52.441554Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amir Feder, Gal Yona, Jonathan Herzig, Matan Eyal, Roee Aharoni, Roi Reichart, Zorik Gekhman","submitted_at":"2024-05-09T17:00:22Z","abstract_excerpt":"When large language models are aligned via supervised fine-tuning, they may encounter new factual information that was not acquired through pre-training. It is often conjectured that this can teach the model the behavior of hallucinating factually incorrect responses, as the model is trained to generate facts that are not grounded in its pre-existing knowledge. In this work, we study the impact of such exposure to new knowledge on the capability of the fine-tuned model to utilize its pre-existing knowledge. To this end, we design a controlled setup, focused on closed-book QA, where we vary the"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.05904","kind":"arxiv","version":3},"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/2405.05904/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":"2405.05904","created_at":"2026-07-05T09:13:52.441617+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.05904v3","created_at":"2026-07-05T09:13:52.441617+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.05904","created_at":"2026-07-05T09:13:52.441617+00:00"},{"alias_kind":"pith_short_12","alias_value":"QQJX4ADBWGK7","created_at":"2026-07-05T09:13:52.441617+00:00"},{"alias_kind":"pith_short_16","alias_value":"QQJX4ADBWGK7N2AV","created_at":"2026-07-05T09:13:52.441617+00:00"},{"alias_kind":"pith_short_8","alias_value":"QQJX4ADB","created_at":"2026-07-05T09:13:52.441617+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.25062","citing_title":"Cultivating Machine Intelligence: The OMEGA Shift from Top-Down Optimization to Autopoietic Cognitive Ecologies","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2506.00886","citing_title":"Position: Agent Should Invoke External Tools ONLY When Epistemically Necessary","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2511.02626","citing_title":"Understanding New-Knowledge-Induced Factual Hallucinations in LLMs: Analysis and Interpretation","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2511.20233","citing_title":"REFLEX: Self-Refining Explainable Fact-Checking via Verdict-Anchored Style Control","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2311.05232","citing_title":"A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions","ref_index":101,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25149","citing_title":"Semantic Layers for Reliable LLM-Powered Data Analytics: A Paired Benchmark of Accuracy and Hallucination Across Three Frontier Models","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15574","citing_title":"Why Fine-Tuning Encourages Hallucinations and How to Fix It","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17458","citing_title":"EHRAG: Bridging Semantic Gaps in Lightweight GraphRAG via Hybrid Hypergraph Construction and Retrieval","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QQJX4ADBWGK7N2AVXVOLLMJGXG","json":"https://pith.science/pith/QQJX4ADBWGK7N2AVXVOLLMJGXG.json","graph_json":"https://pith.science/api/pith-number/QQJX4ADBWGK7N2AVXVOLLMJGXG/graph.json","events_json":"https://pith.science/api/pith-number/QQJX4ADBWGK7N2AVXVOLLMJGXG/events.json","paper":"https://pith.science/paper/QQJX4ADB"},"agent_actions":{"view_html":"https://pith.science/pith/QQJX4ADBWGK7N2AVXVOLLMJGXG","download_json":"https://pith.science/pith/QQJX4ADBWGK7N2AVXVOLLMJGXG.json","view_paper":"https://pith.science/paper/QQJX4ADB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.05904&json=true","fetch_graph":"https://pith.science/api/pith-number/QQJX4ADBWGK7N2AVXVOLLMJGXG/graph.json","fetch_events":"https://pith.science/api/pith-number/QQJX4ADBWGK7N2AVXVOLLMJGXG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QQJX4ADBWGK7N2AVXVOLLMJGXG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QQJX4ADBWGK7N2AVXVOLLMJGXG/action/storage_attestation","attest_author":"https://pith.science/pith/QQJX4ADBWGK7N2AVXVOLLMJGXG/action/author_attestation","sign_citation":"https://pith.science/pith/QQJX4ADBWGK7N2AVXVOLLMJGXG/action/citation_signature","submit_replication":"https://pith.science/pith/QQJX4ADBWGK7N2AVXVOLLMJGXG/action/replication_record"}},"created_at":"2026-07-05T09:13:52.441617+00:00","updated_at":"2026-07-05T09:13:52.441617+00:00"}