{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:TUPS347FM77TYXD2XJI3DWF3BS","short_pith_number":"pith:TUPS347F","schema_version":"1.0","canonical_sha256":"9d1f2df3e567ff3c5c7aba51b1d8bb0ca649961dc22658cf27e35b90fd0b8891","source":{"kind":"arxiv","id":"2504.13227","version":2},"attestation_state":"computed","paper":{"title":"DIDS: Domain Impact-aware Data Sampling for Large Language Model Training","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Hao Chen, Jiajie Xu, Jia Zhu, Jingzhi Fang, Jipeng Zhang, Ruiyuan Zhang, Sirui Han, Weijie Shi, Xiaofang Zhou, Yaguang Wu, Yao Zhao","submitted_at":"2025-04-17T13:09:38Z","abstract_excerpt":"Large language models (LLMs) are commonly trained on multi-domain datasets, where domain sampling strategies significantly impact model performance due to varying domain importance across downstream tasks. Existing approaches for optimizing domain-level sampling strategies struggle with maintaining intra-domain consistency and accurately measuring domain impact. In this paper, we present Domain Impact-aware Data Sampling (DIDS). To ensure intra-domain consistency, a gradient clustering algorithm is proposed to group training data based on their learning effects, where a proxy language model an"},"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":"2504.13227","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-17T13:09:38Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"9fb377428628c218e013c91e278d8f3968212bcdb378eaab8e0d9783c31b540b","abstract_canon_sha256":"e73bab70f2e36d041016847ddf970e608bf191d8e217a09b3b716bbf6e557fc6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:57:36.077152Z","signature_b64":"p0ksfb1CMeeBQeDQur3lkq99LV7kb1Y5iBhgAYNRpCWHCFXpKSUttLBrnFbNYHd9EMLR7irTeyeoSitRYSK2Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9d1f2df3e567ff3c5c7aba51b1d8bb0ca649961dc22658cf27e35b90fd0b8891","last_reissued_at":"2026-07-05T11:57:36.076658Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:57:36.076658Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DIDS: Domain Impact-aware Data Sampling for Large Language Model Training","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Hao Chen, Jiajie Xu, Jia Zhu, Jingzhi Fang, Jipeng Zhang, Ruiyuan Zhang, Sirui Han, Weijie Shi, Xiaofang Zhou, Yaguang Wu, Yao Zhao","submitted_at":"2025-04-17T13:09:38Z","abstract_excerpt":"Large language models (LLMs) are commonly trained on multi-domain datasets, where domain sampling strategies significantly impact model performance due to varying domain importance across downstream tasks. Existing approaches for optimizing domain-level sampling strategies struggle with maintaining intra-domain consistency and accurately measuring domain impact. In this paper, we present Domain Impact-aware Data Sampling (DIDS). To ensure intra-domain consistency, a gradient clustering algorithm is proposed to group training data based on their learning effects, where a proxy language model an"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.13227","kind":"arxiv","version":2},"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/2504.13227/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":"2504.13227","created_at":"2026-07-05T11:57:36.076720+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.13227v2","created_at":"2026-07-05T11:57:36.076720+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.13227","created_at":"2026-07-05T11:57:36.076720+00:00"},{"alias_kind":"pith_short_12","alias_value":"TUPS347FM77T","created_at":"2026-07-05T11:57:36.076720+00:00"},{"alias_kind":"pith_short_16","alias_value":"TUPS347FM77TYXD2","created_at":"2026-07-05T11:57:36.076720+00:00"},{"alias_kind":"pith_short_8","alias_value":"TUPS347F","created_at":"2026-07-05T11:57:36.076720+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/TUPS347FM77TYXD2XJI3DWF3BS","json":"https://pith.science/pith/TUPS347FM77TYXD2XJI3DWF3BS.json","graph_json":"https://pith.science/api/pith-number/TUPS347FM77TYXD2XJI3DWF3BS/graph.json","events_json":"https://pith.science/api/pith-number/TUPS347FM77TYXD2XJI3DWF3BS/events.json","paper":"https://pith.science/paper/TUPS347F"},"agent_actions":{"view_html":"https://pith.science/pith/TUPS347FM77TYXD2XJI3DWF3BS","download_json":"https://pith.science/pith/TUPS347FM77TYXD2XJI3DWF3BS.json","view_paper":"https://pith.science/paper/TUPS347F","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.13227&json=true","fetch_graph":"https://pith.science/api/pith-number/TUPS347FM77TYXD2XJI3DWF3BS/graph.json","fetch_events":"https://pith.science/api/pith-number/TUPS347FM77TYXD2XJI3DWF3BS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TUPS347FM77TYXD2XJI3DWF3BS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TUPS347FM77TYXD2XJI3DWF3BS/action/storage_attestation","attest_author":"https://pith.science/pith/TUPS347FM77TYXD2XJI3DWF3BS/action/author_attestation","sign_citation":"https://pith.science/pith/TUPS347FM77TYXD2XJI3DWF3BS/action/citation_signature","submit_replication":"https://pith.science/pith/TUPS347FM77TYXD2XJI3DWF3BS/action/replication_record"}},"created_at":"2026-07-05T11:57:36.076720+00:00","updated_at":"2026-07-05T11:57:36.076720+00:00"}