{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:6JV3NVMQACXYETKIQYYNQMC5LR","short_pith_number":"pith:6JV3NVMQ","schema_version":"1.0","canonical_sha256":"f26bb6d59000af824d488630d8305d5c4e6ca403ccfb77b208962ab5491a992a","source":{"kind":"arxiv","id":"2504.20644","version":1},"attestation_state":"computed","paper":{"title":"Combatting Dimensional Collapse in LLM Pre-Training Data via Diversified File Selection","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dacheng Tao, Li Shen, Pingjie Wang, Shengchao Hu, Siyuan Du, Yanfeng Wang, Ya Zhang, Ziqing Fan","submitted_at":"2025-04-29T11:13:18Z","abstract_excerpt":"Selecting high-quality pre-training data for large language models (LLMs) is crucial for enhancing their overall performance under limited computation budget, improving both training and sample efficiency. Recent advancements in file selection primarily rely on using an existing or trained proxy model to assess the similarity of samples to a target domain, such as high quality sources BookCorpus and Wikipedia. However, upon revisiting these methods, the domain-similarity selection criteria demonstrates a diversity dilemma, i.e.dimensional collapse in the feature space, improving performance on"},"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":"2504.20644","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-04-29T11:13:18Z","cross_cats_sorted":[],"title_canon_sha256":"03a226c1d491bfe1f3b8d9662fcb88a605bece9af9f9c36568524f8967d49d70","abstract_canon_sha256":"643fd38789dd7602c7de4247fee4d1440c34f28e4dcbc9e173711dc244c1e187"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:55:46.217333Z","signature_b64":"R4fpMUg6LiSuow0q9nw3a0KYBYgI8PIcjpadjfkx11DxLy07OYStRiHQy6Ux3zjDJQkl0CXr0ot2wJ0qf7sSDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f26bb6d59000af824d488630d8305d5c4e6ca403ccfb77b208962ab5491a992a","last_reissued_at":"2026-07-05T10:55:46.216448Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:55:46.216448Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Combatting Dimensional Collapse in LLM Pre-Training Data via Diversified File Selection","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dacheng Tao, Li Shen, Pingjie Wang, Shengchao Hu, Siyuan Du, Yanfeng Wang, Ya Zhang, Ziqing Fan","submitted_at":"2025-04-29T11:13:18Z","abstract_excerpt":"Selecting high-quality pre-training data for large language models (LLMs) is crucial for enhancing their overall performance under limited computation budget, improving both training and sample efficiency. Recent advancements in file selection primarily rely on using an existing or trained proxy model to assess the similarity of samples to a target domain, such as high quality sources BookCorpus and Wikipedia. However, upon revisiting these methods, the domain-similarity selection criteria demonstrates a diversity dilemma, i.e.dimensional collapse in the feature space, improving performance on"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.20644","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/2504.20644/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":"2504.20644","created_at":"2026-07-05T10:55:46.216568+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.20644v1","created_at":"2026-07-05T10:55:46.216568+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.20644","created_at":"2026-07-05T10:55:46.216568+00:00"},{"alias_kind":"pith_short_12","alias_value":"6JV3NVMQACXY","created_at":"2026-07-05T10:55:46.216568+00:00"},{"alias_kind":"pith_short_16","alias_value":"6JV3NVMQACXYETKI","created_at":"2026-07-05T10:55:46.216568+00:00"},{"alias_kind":"pith_short_8","alias_value":"6JV3NVMQ","created_at":"2026-07-05T10:55:46.216568+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.12762","citing_title":"IDEAL: Data Equilibrium Adaptation for Multi-Capability Language Model Alignment","ref_index":17,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6JV3NVMQACXYETKIQYYNQMC5LR","json":"https://pith.science/pith/6JV3NVMQACXYETKIQYYNQMC5LR.json","graph_json":"https://pith.science/api/pith-number/6JV3NVMQACXYETKIQYYNQMC5LR/graph.json","events_json":"https://pith.science/api/pith-number/6JV3NVMQACXYETKIQYYNQMC5LR/events.json","paper":"https://pith.science/paper/6JV3NVMQ"},"agent_actions":{"view_html":"https://pith.science/pith/6JV3NVMQACXYETKIQYYNQMC5LR","download_json":"https://pith.science/pith/6JV3NVMQACXYETKIQYYNQMC5LR.json","view_paper":"https://pith.science/paper/6JV3NVMQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.20644&json=true","fetch_graph":"https://pith.science/api/pith-number/6JV3NVMQACXYETKIQYYNQMC5LR/graph.json","fetch_events":"https://pith.science/api/pith-number/6JV3NVMQACXYETKIQYYNQMC5LR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6JV3NVMQACXYETKIQYYNQMC5LR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6JV3NVMQACXYETKIQYYNQMC5LR/action/storage_attestation","attest_author":"https://pith.science/pith/6JV3NVMQACXYETKIQYYNQMC5LR/action/author_attestation","sign_citation":"https://pith.science/pith/6JV3NVMQACXYETKIQYYNQMC5LR/action/citation_signature","submit_replication":"https://pith.science/pith/6JV3NVMQACXYETKIQYYNQMC5LR/action/replication_record"}},"created_at":"2026-07-05T10:55:46.216568+00:00","updated_at":"2026-07-05T10:55:46.216568+00:00"}