{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KPHTAZ7SRAIDAZVB6TT3CJKYPW","short_pith_number":"pith:KPHTAZ7S","schema_version":"1.0","canonical_sha256":"53cf3067f288103066a1f4e7b125587da7a00235e6f5254ba5d1391eba56eb47","source":{"kind":"arxiv","id":"2603.25120","version":1},"attestation_state":"computed","paper":{"title":"DFLOP: A Data-driven Framework for Multimodal LLM Training Pipeline Optimization","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Chaerim Lim, DongWook Kim, Hyeonjun An, Hyeonsoo Lee, Hyunjoon Kim, Jinkyu Jeong, Kwanghyun Park, Rathijit Sen, Sangmin Jung, Sihyun Kim, Takki Yu, Youngsok Kim","submitted_at":"2026-03-26T07:45:29Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) have achieved remarkable advances by integrating text, image, and audio understanding within a unified architecture. However, existing distributed training frameworks remain fundamentally data-blind: they parallelize computation without accounting for variations in input data characteristics. This data unawareness leads to severe computation skew across stages and microbatches, where heterogeneous multimodal inputs incur different processing costs. Consequently, GPU resources are unevenly utilized, synchronization delays accumulate, and overall training"},"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":"2603.25120","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.DC","submitted_at":"2026-03-26T07:45:29Z","cross_cats_sorted":[],"title_canon_sha256":"ed581e2048e8a391d1249361b99260d1c3ae81dc05ea36a5a4c7f699c6b124c7","abstract_canon_sha256":"e42aa65f0c34b51fc1302b52e237a564ac145905a06ce6fce5f8366f2cade348"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:05:11.815021Z","signature_b64":"1v0L7DXE5/I5R9yyxmOQwvZB7CB5Y2b9E/dBUr990BcMKUU4vf6q5wvPiGq4iyZOYknOay4XzzxZGMqRM0N8DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"53cf3067f288103066a1f4e7b125587da7a00235e6f5254ba5d1391eba56eb47","last_reissued_at":"2026-05-20T01:05:11.814279Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:05:11.814279Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DFLOP: A Data-driven Framework for Multimodal LLM Training Pipeline Optimization","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Chaerim Lim, DongWook Kim, Hyeonjun An, Hyeonsoo Lee, Hyunjoon Kim, Jinkyu Jeong, Kwanghyun Park, Rathijit Sen, Sangmin Jung, Sihyun Kim, Takki Yu, Youngsok Kim","submitted_at":"2026-03-26T07:45:29Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) have achieved remarkable advances by integrating text, image, and audio understanding within a unified architecture. However, existing distributed training frameworks remain fundamentally data-blind: they parallelize computation without accounting for variations in input data characteristics. This data unawareness leads to severe computation skew across stages and microbatches, where heterogeneous multimodal inputs incur different processing costs. Consequently, GPU resources are unevenly utilized, synchronization delays accumulate, and overall training"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.25120","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/2603.25120/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":"2603.25120","created_at":"2026-05-20T01:05:11.814402+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.25120v1","created_at":"2026-05-20T01:05:11.814402+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.25120","created_at":"2026-05-20T01:05:11.814402+00:00"},{"alias_kind":"pith_short_12","alias_value":"KPHTAZ7SRAID","created_at":"2026-05-20T01:05:11.814402+00:00"},{"alias_kind":"pith_short_16","alias_value":"KPHTAZ7SRAIDAZVB","created_at":"2026-05-20T01:05:11.814402+00:00"},{"alias_kind":"pith_short_8","alias_value":"KPHTAZ7S","created_at":"2026-05-20T01:05:11.814402+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/KPHTAZ7SRAIDAZVB6TT3CJKYPW","json":"https://pith.science/pith/KPHTAZ7SRAIDAZVB6TT3CJKYPW.json","graph_json":"https://pith.science/api/pith-number/KPHTAZ7SRAIDAZVB6TT3CJKYPW/graph.json","events_json":"https://pith.science/api/pith-number/KPHTAZ7SRAIDAZVB6TT3CJKYPW/events.json","paper":"https://pith.science/paper/KPHTAZ7S"},"agent_actions":{"view_html":"https://pith.science/pith/KPHTAZ7SRAIDAZVB6TT3CJKYPW","download_json":"https://pith.science/pith/KPHTAZ7SRAIDAZVB6TT3CJKYPW.json","view_paper":"https://pith.science/paper/KPHTAZ7S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.25120&json=true","fetch_graph":"https://pith.science/api/pith-number/KPHTAZ7SRAIDAZVB6TT3CJKYPW/graph.json","fetch_events":"https://pith.science/api/pith-number/KPHTAZ7SRAIDAZVB6TT3CJKYPW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KPHTAZ7SRAIDAZVB6TT3CJKYPW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KPHTAZ7SRAIDAZVB6TT3CJKYPW/action/storage_attestation","attest_author":"https://pith.science/pith/KPHTAZ7SRAIDAZVB6TT3CJKYPW/action/author_attestation","sign_citation":"https://pith.science/pith/KPHTAZ7SRAIDAZVB6TT3CJKYPW/action/citation_signature","submit_replication":"https://pith.science/pith/KPHTAZ7SRAIDAZVB6TT3CJKYPW/action/replication_record"}},"created_at":"2026-05-20T01:05:11.814402+00:00","updated_at":"2026-05-20T01:05:11.814402+00:00"}