{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:Y3HD7TB4JS4NZXLIGXCRN7U37I","short_pith_number":"pith:Y3HD7TB4","schema_version":"1.0","canonical_sha256":"c6ce3fcc3c4cb8dcdd6835c516fe9bfa370fa93bf4be4dbe31414571044dfc49","source":{"kind":"arxiv","id":"2508.17677","version":1},"attestation_state":"computed","paper":{"title":"TiKMiX: Take Data Influence into Dynamic Mixture for Language Model Pre-training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Binbin Liu, Fengze Liu, Jiyao Deng, Taifeng Wang, Weidong Zhou, Xiaohuan Zhou, Xuecheng Wu, Yifan Wang, Yuanfan Guo","submitted_at":"2025-08-25T05:18:32Z","abstract_excerpt":"The data mixture used in the pre-training of a language model is a cornerstone of its final performance. However, a static mixing strategy is suboptimal, as the model's learning preferences for various data domains shift dynamically throughout training. Crucially, observing these evolving preferences in a computationally efficient manner remains a significant challenge. To address this, we propose TiKMiX, a method that dynamically adjusts the data mixture according to the model's evolving preferences. TiKMiX introduces Group Influence, an efficient metric for evaluating the impact of data doma"},"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":"2508.17677","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-25T05:18:32Z","cross_cats_sorted":[],"title_canon_sha256":"9f0c50ffa328a4d993ef8fad9be5526350dca798b203fbcb0266fe8c10eb9e64","abstract_canon_sha256":"f90229af4eae3e4a46ad91d432ccb64ddac5c207c261ce16c4227ebf4f5b74e3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:58:45.656528Z","signature_b64":"7D7Ocu0UhFzhxLUwy9GUlXQnfbk5/FnlNacz2VDv/GWnAvstmh70MVxcs6GLglGU6txT2Np/gbKIOdUf5EyJDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c6ce3fcc3c4cb8dcdd6835c516fe9bfa370fa93bf4be4dbe31414571044dfc49","last_reissued_at":"2026-07-05T11:58:45.656044Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:58:45.656044Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TiKMiX: Take Data Influence into Dynamic Mixture for Language Model Pre-training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Binbin Liu, Fengze Liu, Jiyao Deng, Taifeng Wang, Weidong Zhou, Xiaohuan Zhou, Xuecheng Wu, Yifan Wang, Yuanfan Guo","submitted_at":"2025-08-25T05:18:32Z","abstract_excerpt":"The data mixture used in the pre-training of a language model is a cornerstone of its final performance. However, a static mixing strategy is suboptimal, as the model's learning preferences for various data domains shift dynamically throughout training. Crucially, observing these evolving preferences in a computationally efficient manner remains a significant challenge. To address this, we propose TiKMiX, a method that dynamically adjusts the data mixture according to the model's evolving preferences. TiKMiX introduces Group Influence, an efficient metric for evaluating the impact of data doma"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.17677","kind":"arxiv","version":1},"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/2508.17677/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":"2508.17677","created_at":"2026-07-05T11:58:45.656105+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.17677v1","created_at":"2026-07-05T11:58:45.656105+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.17677","created_at":"2026-07-05T11:58:45.656105+00:00"},{"alias_kind":"pith_short_12","alias_value":"Y3HD7TB4JS4N","created_at":"2026-07-05T11:58:45.656105+00:00"},{"alias_kind":"pith_short_16","alias_value":"Y3HD7TB4JS4NZXLI","created_at":"2026-07-05T11:58:45.656105+00:00"},{"alias_kind":"pith_short_8","alias_value":"Y3HD7TB4","created_at":"2026-07-05T11:58:45.656105+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18663","citing_title":"RegMix-D: Dynamic Data Mixing via Proxy Training Trajectories","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11499","citing_title":"Hubs or Fringes: Pretraining Data Selection via Web Graph Centrality","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16380","citing_title":"Data Mixing for Large Language Models Pretraining: A Survey and Outlook","ref_index":35,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Y3HD7TB4JS4NZXLIGXCRN7U37I","json":"https://pith.science/pith/Y3HD7TB4JS4NZXLIGXCRN7U37I.json","graph_json":"https://pith.science/api/pith-number/Y3HD7TB4JS4NZXLIGXCRN7U37I/graph.json","events_json":"https://pith.science/api/pith-number/Y3HD7TB4JS4NZXLIGXCRN7U37I/events.json","paper":"https://pith.science/paper/Y3HD7TB4"},"agent_actions":{"view_html":"https://pith.science/pith/Y3HD7TB4JS4NZXLIGXCRN7U37I","download_json":"https://pith.science/pith/Y3HD7TB4JS4NZXLIGXCRN7U37I.json","view_paper":"https://pith.science/paper/Y3HD7TB4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.17677&json=true","fetch_graph":"https://pith.science/api/pith-number/Y3HD7TB4JS4NZXLIGXCRN7U37I/graph.json","fetch_events":"https://pith.science/api/pith-number/Y3HD7TB4JS4NZXLIGXCRN7U37I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Y3HD7TB4JS4NZXLIGXCRN7U37I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Y3HD7TB4JS4NZXLIGXCRN7U37I/action/storage_attestation","attest_author":"https://pith.science/pith/Y3HD7TB4JS4NZXLIGXCRN7U37I/action/author_attestation","sign_citation":"https://pith.science/pith/Y3HD7TB4JS4NZXLIGXCRN7U37I/action/citation_signature","submit_replication":"https://pith.science/pith/Y3HD7TB4JS4NZXLIGXCRN7U37I/action/replication_record"}},"created_at":"2026-07-05T11:58:45.656105+00:00","updated_at":"2026-07-05T11:58:45.656105+00:00"}