{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:CMW2D6KRSBYXQ5YVCY5LTPVTIH","short_pith_number":"pith:CMW2D6KR","schema_version":"1.0","canonical_sha256":"132da1f9519071787715163ab9beb341d70dcc2615c58da267504b21f7c2281a","source":{"kind":"arxiv","id":"2509.18150","version":2},"attestation_state":"computed","paper":{"title":"Improving MLLM Training Efficiency via Stage-Aware Sparsity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Baobao Chang, Haozhe Zhao, Kean Shi, Liang Chen","submitted_at":"2025-09-16T11:33:20Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) have demonstrated outstanding performance across a variety of domains. However, training MLLMs is often inefficient, as much of the computation is redundant due to the long input sequences from multimodal data and underutilized inter-layer operations. Notably, such redundancy is not static but varies across different stages of training. Building on this observation, we shift the focus to the training process itself and propose a training-efficient framework based on sparse representations, termed the Sparse Training Scheme (STS). Instead of applying a u"},"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":"2509.18150","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-09-16T11:33:20Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"54d44dd8fadbff181082ea85e3d1f9bc6178149073a2af9840fad749d8ec7700","abstract_canon_sha256":"f2dcddec9b0a99f6ea4771e9633b7af37a1010ccc9cceba16a6e6e564287e610"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:04:13.366020Z","signature_b64":"6Bqu7wWFelW1ifd20m7GqQ8mhG+2lmHzEyJwsG64XGPDBWlaVUGiF3ttYGex5wsQfi4oWEGVyLOsXGabXegZDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"132da1f9519071787715163ab9beb341d70dcc2615c58da267504b21f7c2281a","last_reissued_at":"2026-05-20T00:04:13.365178Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:04:13.365178Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improving MLLM Training Efficiency via Stage-Aware Sparsity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Baobao Chang, Haozhe Zhao, Kean Shi, Liang Chen","submitted_at":"2025-09-16T11:33:20Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) have demonstrated outstanding performance across a variety of domains. However, training MLLMs is often inefficient, as much of the computation is redundant due to the long input sequences from multimodal data and underutilized inter-layer operations. Notably, such redundancy is not static but varies across different stages of training. Building on this observation, we shift the focus to the training process itself and propose a training-efficient framework based on sparse representations, termed the Sparse Training Scheme (STS). Instead of applying a u"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.18150","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/2509.18150/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":"2509.18150","created_at":"2026-05-20T00:04:13.365343+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.18150v2","created_at":"2026-05-20T00:04:13.365343+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.18150","created_at":"2026-05-20T00:04:13.365343+00:00"},{"alias_kind":"pith_short_12","alias_value":"CMW2D6KRSBYX","created_at":"2026-05-20T00:04:13.365343+00:00"},{"alias_kind":"pith_short_16","alias_value":"CMW2D6KRSBYXQ5YV","created_at":"2026-05-20T00:04:13.365343+00:00"},{"alias_kind":"pith_short_8","alias_value":"CMW2D6KR","created_at":"2026-05-20T00:04:13.365343+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2509.18150","citing_title":"Improving MLLM Training Efficiency via Stage-Aware Sparsity","ref_index":2,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CMW2D6KRSBYXQ5YVCY5LTPVTIH","json":"https://pith.science/pith/CMW2D6KRSBYXQ5YVCY5LTPVTIH.json","graph_json":"https://pith.science/api/pith-number/CMW2D6KRSBYXQ5YVCY5LTPVTIH/graph.json","events_json":"https://pith.science/api/pith-number/CMW2D6KRSBYXQ5YVCY5LTPVTIH/events.json","paper":"https://pith.science/paper/CMW2D6KR"},"agent_actions":{"view_html":"https://pith.science/pith/CMW2D6KRSBYXQ5YVCY5LTPVTIH","download_json":"https://pith.science/pith/CMW2D6KRSBYXQ5YVCY5LTPVTIH.json","view_paper":"https://pith.science/paper/CMW2D6KR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.18150&json=true","fetch_graph":"https://pith.science/api/pith-number/CMW2D6KRSBYXQ5YVCY5LTPVTIH/graph.json","fetch_events":"https://pith.science/api/pith-number/CMW2D6KRSBYXQ5YVCY5LTPVTIH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CMW2D6KRSBYXQ5YVCY5LTPVTIH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CMW2D6KRSBYXQ5YVCY5LTPVTIH/action/storage_attestation","attest_author":"https://pith.science/pith/CMW2D6KRSBYXQ5YVCY5LTPVTIH/action/author_attestation","sign_citation":"https://pith.science/pith/CMW2D6KRSBYXQ5YVCY5LTPVTIH/action/citation_signature","submit_replication":"https://pith.science/pith/CMW2D6KRSBYXQ5YVCY5LTPVTIH/action/replication_record"}},"created_at":"2026-05-20T00:04:13.365343+00:00","updated_at":"2026-05-20T00:04:13.365343+00:00"}