{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:2ITIIGRPYV3YSII3QFNX4PFJZE","short_pith_number":"pith:2ITIIGRP","schema_version":"1.0","canonical_sha256":"d226841a2fc57789211b815b7e3ca9c91b1e86a1d15a43d82a60fc18141d4fe5","source":{"kind":"arxiv","id":"2412.06289","version":3},"attestation_state":"computed","paper":{"title":"S$^{2}$FT: Efficient, Scalable and Generalizable LLM Fine-tuning by Structured Sparsity","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Beidi Chen, Geyang Guo, Huaxiu Yao, Jiawei Zhao, Jixuan Leng, Linjun Zhang, Ryumei Nakada, Xinyu Yang","submitted_at":"2024-12-09T08:24:11Z","abstract_excerpt":"Current PEFT methods for LLMs can achieve either high quality, efficient training, or scalable serving, but not all three simultaneously. To address this limitation, we investigate sparse fine-tuning and observe a remarkable improvement in generalization ability. Utilizing this key insight, we propose a family of Structured Sparse Fine-Tuning (S$^{2}$FT) methods for LLMs, which concurrently achieve state-of-the-art fine-tuning performance, training efficiency, and inference scalability. S$^{2}$FT accomplishes this by \"selecting sparsely and computing densely\". It selects a few heads and channe"},"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":"2412.06289","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-12-09T08:24:11Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"ba811d941ba9710f46fb552e2dc862e8e472406b72675784dd5c8bd2ce4ac5aa","abstract_canon_sha256":"4e62c02ae7d42042100535eebcebffee33b3e7cd8ec0373835fb8babab704b7b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:51:39.292244Z","signature_b64":"PanhE64B2Um50Wutk2RkzydY11BI9fO/66UF/B2qZYLRf+cN9ANHybLfIulIjCtA9bt+9IpjkHQGnz5o1+2bDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d226841a2fc57789211b815b7e3ca9c91b1e86a1d15a43d82a60fc18141d4fe5","last_reissued_at":"2026-07-05T09:51:39.291743Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:51:39.291743Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"S$^{2}$FT: Efficient, Scalable and Generalizable LLM Fine-tuning by Structured Sparsity","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Beidi Chen, Geyang Guo, Huaxiu Yao, Jiawei Zhao, Jixuan Leng, Linjun Zhang, Ryumei Nakada, Xinyu Yang","submitted_at":"2024-12-09T08:24:11Z","abstract_excerpt":"Current PEFT methods for LLMs can achieve either high quality, efficient training, or scalable serving, but not all three simultaneously. To address this limitation, we investigate sparse fine-tuning and observe a remarkable improvement in generalization ability. Utilizing this key insight, we propose a family of Structured Sparse Fine-Tuning (S$^{2}$FT) methods for LLMs, which concurrently achieve state-of-the-art fine-tuning performance, training efficiency, and inference scalability. S$^{2}$FT accomplishes this by \"selecting sparsely and computing densely\". It selects a few heads and channe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.06289","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/2412.06289/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":"2412.06289","created_at":"2026-07-05T09:51:39.291811+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.06289v3","created_at":"2026-07-05T09:51:39.291811+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.06289","created_at":"2026-07-05T09:51:39.291811+00:00"},{"alias_kind":"pith_short_12","alias_value":"2ITIIGRPYV3Y","created_at":"2026-07-05T09:51:39.291811+00:00"},{"alias_kind":"pith_short_16","alias_value":"2ITIIGRPYV3YSII3","created_at":"2026-07-05T09:51:39.291811+00:00"},{"alias_kind":"pith_short_8","alias_value":"2ITIIGRP","created_at":"2026-07-05T09:51:39.291811+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/2ITIIGRPYV3YSII3QFNX4PFJZE","json":"https://pith.science/pith/2ITIIGRPYV3YSII3QFNX4PFJZE.json","graph_json":"https://pith.science/api/pith-number/2ITIIGRPYV3YSII3QFNX4PFJZE/graph.json","events_json":"https://pith.science/api/pith-number/2ITIIGRPYV3YSII3QFNX4PFJZE/events.json","paper":"https://pith.science/paper/2ITIIGRP"},"agent_actions":{"view_html":"https://pith.science/pith/2ITIIGRPYV3YSII3QFNX4PFJZE","download_json":"https://pith.science/pith/2ITIIGRPYV3YSII3QFNX4PFJZE.json","view_paper":"https://pith.science/paper/2ITIIGRP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.06289&json=true","fetch_graph":"https://pith.science/api/pith-number/2ITIIGRPYV3YSII3QFNX4PFJZE/graph.json","fetch_events":"https://pith.science/api/pith-number/2ITIIGRPYV3YSII3QFNX4PFJZE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2ITIIGRPYV3YSII3QFNX4PFJZE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2ITIIGRPYV3YSII3QFNX4PFJZE/action/storage_attestation","attest_author":"https://pith.science/pith/2ITIIGRPYV3YSII3QFNX4PFJZE/action/author_attestation","sign_citation":"https://pith.science/pith/2ITIIGRPYV3YSII3QFNX4PFJZE/action/citation_signature","submit_replication":"https://pith.science/pith/2ITIIGRPYV3YSII3QFNX4PFJZE/action/replication_record"}},"created_at":"2026-07-05T09:51:39.291811+00:00","updated_at":"2026-07-05T09:51:39.291811+00:00"}