{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:6OCSUEHKJA7CUQCSSOFDWJIUQT","short_pith_number":"pith:6OCSUEHK","schema_version":"1.0","canonical_sha256":"f3852a10ea483e2a4052938a3b251484ec738672859f889bb2f8a8d8a24e30e9","source":{"kind":"arxiv","id":"2406.12227","version":3},"attestation_state":"computed","paper":{"title":"Refine Large Language Model Fine-tuning via Instruction Vector","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Defu Lian, Gangwei Jiang, Ying Wei, Zhaoyi Li","submitted_at":"2024-06-18T03:05:08Z","abstract_excerpt":"Fine-tuning large language models (LLMs) can cause them to lose their general capabilities. However, the intrinsic mechanisms behind such forgetting remain unexplored. In this paper, we begin by examining this phenomenon by focusing on knowledge understanding and instruction following, with the latter identified as the main contributor to forgetting during fine-tuning. Consequently, we propose the Instruction Vector (IV) framework to capture model representations highly related to specific instruction-following capabilities, thereby making it possible to understand model-intrinsic forgetting. "},"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":"2406.12227","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-06-18T03:05:08Z","cross_cats_sorted":[],"title_canon_sha256":"92369abe6fbc5884f03e2029ebcaf1e7a2e194370ea494521839662d596c60c8","abstract_canon_sha256":"46f82a7db8473beb05bc227a584ec99ccc4cd7d7ddd171faae046347a3e7b046"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:41:38.123815Z","signature_b64":"1Ggu5r8WVDxtiB3B3euqrYJOGMRJpbRNWlfbZmxF1vQM0dP4Rfy44NLn4fVDrnC0kHYBhxw0QjAck93KyyDBCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f3852a10ea483e2a4052938a3b251484ec738672859f889bb2f8a8d8a24e30e9","last_reissued_at":"2026-07-05T09:41:38.123389Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:41:38.123389Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Refine Large Language Model Fine-tuning via Instruction Vector","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Defu Lian, Gangwei Jiang, Ying Wei, Zhaoyi Li","submitted_at":"2024-06-18T03:05:08Z","abstract_excerpt":"Fine-tuning large language models (LLMs) can cause them to lose their general capabilities. However, the intrinsic mechanisms behind such forgetting remain unexplored. In this paper, we begin by examining this phenomenon by focusing on knowledge understanding and instruction following, with the latter identified as the main contributor to forgetting during fine-tuning. Consequently, we propose the Instruction Vector (IV) framework to capture model representations highly related to specific instruction-following capabilities, thereby making it possible to understand model-intrinsic forgetting. "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.12227","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/2406.12227/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":"2406.12227","created_at":"2026-07-05T09:41:38.123446+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.12227v3","created_at":"2026-07-05T09:41:38.123446+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.12227","created_at":"2026-07-05T09:41:38.123446+00:00"},{"alias_kind":"pith_short_12","alias_value":"6OCSUEHKJA7C","created_at":"2026-07-05T09:41:38.123446+00:00"},{"alias_kind":"pith_short_16","alias_value":"6OCSUEHKJA7CUQCS","created_at":"2026-07-05T09:41:38.123446+00:00"},{"alias_kind":"pith_short_8","alias_value":"6OCSUEHK","created_at":"2026-07-05T09:41:38.123446+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.20005","citing_title":"Fine-Tuning Without Forgetting via Loss-Adaptive Learning Rates","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2601.03938","citing_title":"FOREVER: Forgetting Curve-Inspired Memory Replay for Language Model Continual Learning","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2601.14004","citing_title":"Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models","ref_index":132,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6OCSUEHKJA7CUQCSSOFDWJIUQT","json":"https://pith.science/pith/6OCSUEHKJA7CUQCSSOFDWJIUQT.json","graph_json":"https://pith.science/api/pith-number/6OCSUEHKJA7CUQCSSOFDWJIUQT/graph.json","events_json":"https://pith.science/api/pith-number/6OCSUEHKJA7CUQCSSOFDWJIUQT/events.json","paper":"https://pith.science/paper/6OCSUEHK"},"agent_actions":{"view_html":"https://pith.science/pith/6OCSUEHKJA7CUQCSSOFDWJIUQT","download_json":"https://pith.science/pith/6OCSUEHKJA7CUQCSSOFDWJIUQT.json","view_paper":"https://pith.science/paper/6OCSUEHK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.12227&json=true","fetch_graph":"https://pith.science/api/pith-number/6OCSUEHKJA7CUQCSSOFDWJIUQT/graph.json","fetch_events":"https://pith.science/api/pith-number/6OCSUEHKJA7CUQCSSOFDWJIUQT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6OCSUEHKJA7CUQCSSOFDWJIUQT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6OCSUEHKJA7CUQCSSOFDWJIUQT/action/storage_attestation","attest_author":"https://pith.science/pith/6OCSUEHKJA7CUQCSSOFDWJIUQT/action/author_attestation","sign_citation":"https://pith.science/pith/6OCSUEHKJA7CUQCSSOFDWJIUQT/action/citation_signature","submit_replication":"https://pith.science/pith/6OCSUEHKJA7CUQCSSOFDWJIUQT/action/replication_record"}},"created_at":"2026-07-05T09:41:38.123446+00:00","updated_at":"2026-07-05T09:41:38.123446+00:00"}