{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:36ZQ3423D2BRPDBFMQFULLY4BX","short_pith_number":"pith:36ZQ3423","schema_version":"1.0","canonical_sha256":"dfb30df35b1e83178c25640b45af1c0de01944bad7b5a800419741b7bad5c1d5","source":{"kind":"arxiv","id":"2402.18243","version":3},"attestation_state":"computed","paper":{"title":"Learning or Self-aligning? Rethinking Instruction Fine-tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Boxi Cao, Cao Liu, Guanglu Wan, Hongyu Lin, Ke Zeng, Le Sun, Mengjie Ren, Xianpei Han, Xunliang Cai","submitted_at":"2024-02-28T11:16:00Z","abstract_excerpt":"Instruction Fine-tuning~(IFT) is a critical phase in building large language models~(LLMs). Previous works mainly focus on the IFT's role in the transfer of behavioral norms and the learning of additional world knowledge. However, the understanding of the underlying mechanisms of IFT remains significantly limited. In this paper, we design a knowledge intervention framework to decouple the potential underlying factors of IFT, thereby enabling individual analysis of different factors. Surprisingly, our experiments reveal that attempting to learn additional world knowledge through IFT often strug"},"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":"2402.18243","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-02-28T11:16:00Z","cross_cats_sorted":[],"title_canon_sha256":"b54f6217c3df95cf2aecdb706102a2e26103913fb02e3c57ff84a905ccd5ae84","abstract_canon_sha256":"4b35ad2dbffdd1603aef1ce1223c9807948f4cba9e66f2368492acbde7e981c7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:54:07.608258Z","signature_b64":"B5M1di6og70fZVCrgwI2c/U7gMWU4kAHF2Z6BfFbcMHOjUp3gwbpdtyaz4u1GsK7DkjjDUA+HuUAhsGMD/8HDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dfb30df35b1e83178c25640b45af1c0de01944bad7b5a800419741b7bad5c1d5","last_reissued_at":"2026-07-05T08:54:07.607747Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:54:07.607747Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning or Self-aligning? Rethinking Instruction Fine-tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Boxi Cao, Cao Liu, Guanglu Wan, Hongyu Lin, Ke Zeng, Le Sun, Mengjie Ren, Xianpei Han, Xunliang Cai","submitted_at":"2024-02-28T11:16:00Z","abstract_excerpt":"Instruction Fine-tuning~(IFT) is a critical phase in building large language models~(LLMs). Previous works mainly focus on the IFT's role in the transfer of behavioral norms and the learning of additional world knowledge. However, the understanding of the underlying mechanisms of IFT remains significantly limited. In this paper, we design a knowledge intervention framework to decouple the potential underlying factors of IFT, thereby enabling individual analysis of different factors. Surprisingly, our experiments reveal that attempting to learn additional world knowledge through IFT often strug"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.18243","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/2402.18243/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":"2402.18243","created_at":"2026-07-05T08:54:07.607806+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.18243v3","created_at":"2026-07-05T08:54:07.607806+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.18243","created_at":"2026-07-05T08:54:07.607806+00:00"},{"alias_kind":"pith_short_12","alias_value":"36ZQ3423D2BR","created_at":"2026-07-05T08:54:07.607806+00:00"},{"alias_kind":"pith_short_16","alias_value":"36ZQ3423D2BRPDBF","created_at":"2026-07-05T08:54:07.607806+00:00"},{"alias_kind":"pith_short_8","alias_value":"36ZQ3423","created_at":"2026-07-05T08:54:07.607806+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.22375","citing_title":"Pangu Embedded: An Efficient Dual-system LLM Reasoner with Metacognition","ref_index":35,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/36ZQ3423D2BRPDBFMQFULLY4BX","json":"https://pith.science/pith/36ZQ3423D2BRPDBFMQFULLY4BX.json","graph_json":"https://pith.science/api/pith-number/36ZQ3423D2BRPDBFMQFULLY4BX/graph.json","events_json":"https://pith.science/api/pith-number/36ZQ3423D2BRPDBFMQFULLY4BX/events.json","paper":"https://pith.science/paper/36ZQ3423"},"agent_actions":{"view_html":"https://pith.science/pith/36ZQ3423D2BRPDBFMQFULLY4BX","download_json":"https://pith.science/pith/36ZQ3423D2BRPDBFMQFULLY4BX.json","view_paper":"https://pith.science/paper/36ZQ3423","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.18243&json=true","fetch_graph":"https://pith.science/api/pith-number/36ZQ3423D2BRPDBFMQFULLY4BX/graph.json","fetch_events":"https://pith.science/api/pith-number/36ZQ3423D2BRPDBFMQFULLY4BX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/36ZQ3423D2BRPDBFMQFULLY4BX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/36ZQ3423D2BRPDBFMQFULLY4BX/action/storage_attestation","attest_author":"https://pith.science/pith/36ZQ3423D2BRPDBFMQFULLY4BX/action/author_attestation","sign_citation":"https://pith.science/pith/36ZQ3423D2BRPDBFMQFULLY4BX/action/citation_signature","submit_replication":"https://pith.science/pith/36ZQ3423D2BRPDBFMQFULLY4BX/action/replication_record"}},"created_at":"2026-07-05T08:54:07.607806+00:00","updated_at":"2026-07-05T08:54:07.607806+00:00"}