{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:3CF6TPWUXOC2BJKUVWBF2JKX26","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"f90620ab9f73f6ef880cec6e7db56e159c3020357f0c959f11f48893daa59db5","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-08T21:46:16Z","title_canon_sha256":"1b5148e548e6b4015b9c36280d09946883ff2a59ef212dc0770c29192d5c4b8c"},"schema_version":"1.0","source":{"id":"2604.00001","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.00001","created_at":"2026-05-18T03:10:03Z"},{"alias_kind":"arxiv_version","alias_value":"2604.00001v2","created_at":"2026-05-18T03:10:03Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.00001","created_at":"2026-05-18T03:10:03Z"},{"alias_kind":"pith_short_12","alias_value":"3CF6TPWUXOC2","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"3CF6TPWUXOC2BJKU","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"3CF6TPWU","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:e96b07b6a604eaff12964b5cf070075f39c7e616f0f9c8a2f22090f8f4d1e943","target":"graph","created_at":"2026-05-18T03:10:03Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Experiments show that our method consistently improves convergence and downstream performance over existing online data selection baselines under the same data budget."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the optimizer-aware update-matching formulation correctly captures sample utility and that the two-stage filter-plus-weight procedure can be computed efficiently without introducing new biases for long-context LLM data."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Filter-then-Weight is a two-stage optimizer-aware method that filters geometrically useful data candidates and optimizes their coefficients to shape target updates in online LLM fine-tuning."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"An optimizer-aware Filter-then-Weight method improves convergence in online LLM fine-tuning by matching updates to the current optimizer state."}],"snapshot_sha256":"fa3fc5ae443ef7236f6e1a4aafa1e60424be077bd397cd9b0c862e1aa4bd361f"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Gradient-based data selection offers a principled framework for estimating sample utility in large language model (LLM) fine-tuning, but existing methods are mostly designed for offline settings. They are therefore less suited to online fine-tuning, where data arrives sequentially, sample utility is step-dependent, and the effective update geometry is shaped by adaptive optimizers. We propose an optimizer-aware framework for gradient-based online data selection and reweighting in LLM fine-tuning. Our key idea is to view online selection not as static sample ranking, but as shaping the next tar","authors_text":"Fangxin Wang, Henry Peng Zou, Langzhou He, Peyman Baghershahi, Philip S. Yu, Sourav Medya","cross_cats":["cs.AI","cs.CL"],"headline":"An optimizer-aware Filter-then-Weight method improves convergence in online LLM fine-tuning by matching updates to the current optimizer state.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-08T21:46:16Z","title":"Filter-then-Weight: Online Data Selection and Reweighting for LLM Fine-Tuning"},"references":{"count":5,"internal_anchors":0,"resolved_work":5,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"arXiv preprint arXiv:2308.03296 , year=","work_id":"0d856ab3-d64d-4427-918f-344d35f6818d","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"From quantity to quality: Boosting llm performance with self-guided data selection for instruction tuning","work_id":"6dd32f47-dcfa-4936-8863-f75c6d55383f","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Yulei Qin, Yuncheng Yang, Pengcheng Guo, Gang Li, Hang Shao, Yuchen Shi, Zihan Xu, Yun Gu, Ke Li, and Xing Sun","work_id":"75dbb3fd-a3d3-4dde-8bcc-72d604931ff9","year":2020},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Tagcos: Task- agnostic gradient clustered coreset selection for instruction tuning data","work_id":"a91d8610-d88d-479a-bab1-63ec127401b0","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"As discussed in Section 4.4, the top-k filtering based can achieve comparable peak performance, with little increased cost compared with filter-only methods","work_id":"98257d07-0500-4fc1-af48-6da2e2073f50","year":2020}],"snapshot_sha256":"3f124652fb96add356163ceb12955937946ee8568aa1119ef1faf7beea98c1f7"},"source":{"id":"2604.00001","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-15T14:20:09.700915Z","id":"317d1bc6-9c6c-4ad9-a3b3-8123976d230e","model_set":{"reader":"grok-4.3"},"one_line_summary":"Filter-then-Weight is a two-stage optimizer-aware method that filters geometrically useful data candidates and optimizes their coefficients to shape target updates in online LLM fine-tuning.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"An optimizer-aware Filter-then-Weight method improves convergence in online LLM fine-tuning by matching updates to the current optimizer state.","strongest_claim":"Experiments show that our method consistently improves convergence and downstream performance over existing online data selection baselines under the same data budget.","weakest_assumption":"That the optimizer-aware update-matching formulation correctly captures sample utility and that the two-stage filter-plus-weight procedure can be computed efficiently without introducing new biases for long-context LLM data."}},"verdict_id":"317d1bc6-9c6c-4ad9-a3b3-8123976d230e"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:e05852303d39a45d8b2c3e1e9d202f05ba27d056b55d9b69617325ff8f240015","target":"record","created_at":"2026-05-18T03:10:03Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"f90620ab9f73f6ef880cec6e7db56e159c3020357f0c959f11f48893daa59db5","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-08T21:46:16Z","title_canon_sha256":"1b5148e548e6b4015b9c36280d09946883ff2a59ef212dc0770c29192d5c4b8c"},"schema_version":"1.0","source":{"id":"2604.00001","kind":"arxiv","version":2}},"canonical_sha256":"d88be9bed4bb85a0a554ad825d2557d7a8cd03f7ac0ff01b41054b5386f7566c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d88be9bed4bb85a0a554ad825d2557d7a8cd03f7ac0ff01b41054b5386f7566c","first_computed_at":"2026-05-18T03:10:03.368690Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:10:03.368690Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"cH7CW083FqlCaSVccpUS3lMAzoIk/cRgtmnd4oGwpu3D6rzbTN9pOQ9A6PcSqrp7X0wpLBKbsWm9IZjjddvTBA==","signature_status":"signed_v1","signed_at":"2026-05-18T03:10:03.369134Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.00001","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e05852303d39a45d8b2c3e1e9d202f05ba27d056b55d9b69617325ff8f240015","sha256:e96b07b6a604eaff12964b5cf070075f39c7e616f0f9c8a2f22090f8f4d1e943"],"state_sha256":"e9ebcd37476ff4975454cefe910cf64062a6792b725d0de180d348bdee5c81eb"}