{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:2Y53Z5MGNLY6UU4VB42EAYFEDG","short_pith_number":"pith:2Y53Z5MG","schema_version":"1.0","canonical_sha256":"d63bbcf5866af1ea53950f344060a419bca3329f47d9200f87aee20cf937d2af","source":{"kind":"arxiv","id":"2409.16986","version":2},"attestation_state":"computed","paper":{"title":"Harnessing Diversity for Important Data Selection in Pretraining Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chengliang Chai, Chi Zhang, Conghui He, Guoren Wang, Huaping Zhong, Jiantao Qiu, Ju Fan, Kuan Zhang, Lei Cao, Rui Wang, Tianyi Bai, Xinlin Zhuang, Ye Yuan","submitted_at":"2024-09-25T14:49:29Z","abstract_excerpt":"Data selection is of great significance in pre-training large language models, given the variation in quality within the large-scale available training corpora. To achieve this, researchers are currently investigating the use of data influence to measure the importance of data instances, $i.e.,$ a high influence score indicates that incorporating this instance to the training set is likely to enhance the model performance. Consequently, they select the top-$k$ instances with the highest scores. However, this approach has several limitations. (1) Computing the influence of all available data is"},"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":"2409.16986","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-09-25T14:49:29Z","cross_cats_sorted":[],"title_canon_sha256":"0cca7305e4d3183a090a1b3e62350e0655be39e060378314e258aeaa4c8989d1","abstract_canon_sha256":"2de6203743b36cb283d0b631ad599958250c3233c203cee14e47025ebe963247"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:16:21.724687Z","signature_b64":"X7eEGijAWy8dUjHCeTzCGtRBndx1vLvi0ocvnBY5eRuzZDA7yT9PjWCNFGXw7JOVfS/hMgMKixUB4W8c45rpDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d63bbcf5866af1ea53950f344060a419bca3329f47d9200f87aee20cf937d2af","last_reissued_at":"2026-07-05T09:16:21.724161Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:16:21.724161Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Harnessing Diversity for Important Data Selection in Pretraining Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chengliang Chai, Chi Zhang, Conghui He, Guoren Wang, Huaping Zhong, Jiantao Qiu, Ju Fan, Kuan Zhang, Lei Cao, Rui Wang, Tianyi Bai, Xinlin Zhuang, Ye Yuan","submitted_at":"2024-09-25T14:49:29Z","abstract_excerpt":"Data selection is of great significance in pre-training large language models, given the variation in quality within the large-scale available training corpora. To achieve this, researchers are currently investigating the use of data influence to measure the importance of data instances, $i.e.,$ a high influence score indicates that incorporating this instance to the training set is likely to enhance the model performance. Consequently, they select the top-$k$ instances with the highest scores. However, this approach has several limitations. (1) Computing the influence of all available data is"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.16986","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/2409.16986/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":"2409.16986","created_at":"2026-07-05T09:16:21.724219+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.16986v2","created_at":"2026-07-05T09:16:21.724219+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.16986","created_at":"2026-07-05T09:16:21.724219+00:00"},{"alias_kind":"pith_short_12","alias_value":"2Y53Z5MGNLY6","created_at":"2026-07-05T09:16:21.724219+00:00"},{"alias_kind":"pith_short_16","alias_value":"2Y53Z5MGNLY6UU4V","created_at":"2026-07-05T09:16:21.724219+00:00"},{"alias_kind":"pith_short_8","alias_value":"2Y53Z5MG","created_at":"2026-07-05T09:16:21.724219+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.11810","citing_title":"GRACE: A Dynamic Coreset Selection Framework for Large Language Model Optimization","ref_index":82,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2Y53Z5MGNLY6UU4VB42EAYFEDG","json":"https://pith.science/pith/2Y53Z5MGNLY6UU4VB42EAYFEDG.json","graph_json":"https://pith.science/api/pith-number/2Y53Z5MGNLY6UU4VB42EAYFEDG/graph.json","events_json":"https://pith.science/api/pith-number/2Y53Z5MGNLY6UU4VB42EAYFEDG/events.json","paper":"https://pith.science/paper/2Y53Z5MG"},"agent_actions":{"view_html":"https://pith.science/pith/2Y53Z5MGNLY6UU4VB42EAYFEDG","download_json":"https://pith.science/pith/2Y53Z5MGNLY6UU4VB42EAYFEDG.json","view_paper":"https://pith.science/paper/2Y53Z5MG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.16986&json=true","fetch_graph":"https://pith.science/api/pith-number/2Y53Z5MGNLY6UU4VB42EAYFEDG/graph.json","fetch_events":"https://pith.science/api/pith-number/2Y53Z5MGNLY6UU4VB42EAYFEDG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2Y53Z5MGNLY6UU4VB42EAYFEDG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2Y53Z5MGNLY6UU4VB42EAYFEDG/action/storage_attestation","attest_author":"https://pith.science/pith/2Y53Z5MGNLY6UU4VB42EAYFEDG/action/author_attestation","sign_citation":"https://pith.science/pith/2Y53Z5MGNLY6UU4VB42EAYFEDG/action/citation_signature","submit_replication":"https://pith.science/pith/2Y53Z5MGNLY6UU4VB42EAYFEDG/action/replication_record"}},"created_at":"2026-07-05T09:16:21.724219+00:00","updated_at":"2026-07-05T09:16:21.724219+00:00"}