{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:SVB3DGHRIIAXLLV6EWC3YJHNRO","short_pith_number":"pith:SVB3DGHR","schema_version":"1.0","canonical_sha256":"9543b198f1420175aebe2585bc24ed8bbf9bf85d8dc812f6680251f3c370d047","source":{"kind":"arxiv","id":"2403.09559","version":4},"attestation_state":"computed","paper":{"title":"Less is More: High-value Data Selection for Visual Instruction Tuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Dawei Gao, Ji-Rong Wen, Kun Zhou, Wayne Xin Zhao, Yaliang Li, Zikang Liu","submitted_at":"2024-03-14T16:47:25Z","abstract_excerpt":"Visual instruction tuning is the key to building large vision language models~(LVLMs), which can greatly improve the task generalization and solving capabilities by learning a mixture of instruction data from diverse visual tasks. Previous work mostly collects multiple existing visual instruction datasets via heuristic ways for training (even more than a million instructions), which may introduce data redundancy and enlarge the training cost. To investigate this issue, we conduct a series of empirical studies, which reveal a significant redundancy within the visual instruction datasets, and sh"},"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":"2403.09559","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-03-14T16:47:25Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"429d83290580d967fdc7f9c63a29df91b690895cbd9e53d5a87996c8089e3025","abstract_canon_sha256":"b3384e9c091e21d3ba80173aaeeb104201cdf73230601686bd67fdabf005c407"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:18:23.775417Z","signature_b64":"ufWDDX5g7P9NM4i8vpwmkAiTT+f4AA5qc2hSQ8pLIT1BERo7ks/SShQ9BCHEZWgIdnYQ5liwKncRr4TpHMLBDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9543b198f1420175aebe2585bc24ed8bbf9bf85d8dc812f6680251f3c370d047","last_reissued_at":"2026-07-05T09:18:23.774916Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:18:23.774916Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Less is More: High-value Data Selection for Visual Instruction Tuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Dawei Gao, Ji-Rong Wen, Kun Zhou, Wayne Xin Zhao, Yaliang Li, Zikang Liu","submitted_at":"2024-03-14T16:47:25Z","abstract_excerpt":"Visual instruction tuning is the key to building large vision language models~(LVLMs), which can greatly improve the task generalization and solving capabilities by learning a mixture of instruction data from diverse visual tasks. Previous work mostly collects multiple existing visual instruction datasets via heuristic ways for training (even more than a million instructions), which may introduce data redundancy and enlarge the training cost. To investigate this issue, we conduct a series of empirical studies, which reveal a significant redundancy within the visual instruction datasets, and sh"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.09559","kind":"arxiv","version":4},"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/2403.09559/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":"2403.09559","created_at":"2026-07-05T09:18:23.774969+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.09559v4","created_at":"2026-07-05T09:18:23.774969+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.09559","created_at":"2026-07-05T09:18:23.774969+00:00"},{"alias_kind":"pith_short_12","alias_value":"SVB3DGHRIIAX","created_at":"2026-07-05T09:18:23.774969+00:00"},{"alias_kind":"pith_short_16","alias_value":"SVB3DGHRIIAXLLV6","created_at":"2026-07-05T09:18:23.774969+00:00"},{"alias_kind":"pith_short_8","alias_value":"SVB3DGHR","created_at":"2026-07-05T09:18:23.774969+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2504.21850","citing_title":"Visual Compositional Tuning","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15691","citing_title":"SEED: Targeted Data Selection by Weighted Independent Set","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13130","citing_title":"GRACE: Gradient-aligned Reasoning Data Curation for Efficient Post-training","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SVB3DGHRIIAXLLV6EWC3YJHNRO","json":"https://pith.science/pith/SVB3DGHRIIAXLLV6EWC3YJHNRO.json","graph_json":"https://pith.science/api/pith-number/SVB3DGHRIIAXLLV6EWC3YJHNRO/graph.json","events_json":"https://pith.science/api/pith-number/SVB3DGHRIIAXLLV6EWC3YJHNRO/events.json","paper":"https://pith.science/paper/SVB3DGHR"},"agent_actions":{"view_html":"https://pith.science/pith/SVB3DGHRIIAXLLV6EWC3YJHNRO","download_json":"https://pith.science/pith/SVB3DGHRIIAXLLV6EWC3YJHNRO.json","view_paper":"https://pith.science/paper/SVB3DGHR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.09559&json=true","fetch_graph":"https://pith.science/api/pith-number/SVB3DGHRIIAXLLV6EWC3YJHNRO/graph.json","fetch_events":"https://pith.science/api/pith-number/SVB3DGHRIIAXLLV6EWC3YJHNRO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SVB3DGHRIIAXLLV6EWC3YJHNRO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SVB3DGHRIIAXLLV6EWC3YJHNRO/action/storage_attestation","attest_author":"https://pith.science/pith/SVB3DGHRIIAXLLV6EWC3YJHNRO/action/author_attestation","sign_citation":"https://pith.science/pith/SVB3DGHRIIAXLLV6EWC3YJHNRO/action/citation_signature","submit_replication":"https://pith.science/pith/SVB3DGHRIIAXLLV6EWC3YJHNRO/action/replication_record"}},"created_at":"2026-07-05T09:18:23.774969+00:00","updated_at":"2026-07-05T09:18:23.774969+00:00"}