{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LSNNIBDSECYZPV4JWPEPMMVSXG","short_pith_number":"pith:LSNNIBDS","schema_version":"1.0","canonical_sha256":"5c9ad4047220b197d789b3c8f632b2b9b13bda066f2f078c324d9e4c8a7b0709","source":{"kind":"arxiv","id":"2501.04155","version":1},"attestation_state":"computed","paper":{"title":"MM-GEN: Enhancing Task Performance Through Targeted Multimodal Data Curation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Baharan Mirzasoleiman, Besmira Nushi, Neel Joshi, Siddharth Joshi, Varun Chandrasekaran, Vibhav Vineet, Vidhisha Balachandran","submitted_at":"2025-01-07T21:55:56Z","abstract_excerpt":"Vision-language models (VLMs) are highly effective but often underperform on specialized tasks; for example, Llava-1.5 struggles with chart and diagram understanding due to scarce task-specific training data. Existing training data, sourced from general-purpose datasets, fails to capture the nuanced details needed for these tasks. We introduce MM-Gen, a scalable method that generates task-specific, high-quality synthetic text for candidate images by leveraging stronger models. MM-Gen employs a three-stage targeted process: partitioning data into subgroups, generating targeted text based on tas"},"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":"2501.04155","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-01-07T21:55:56Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"249701a55ee6cc1b883050b46b91158b00c78937f73c5e37c1c5718a2a5fbc1d","abstract_canon_sha256":"f0c4a8f7aefa492e28142e97dd9bbc5d056aa4a0530a0e1514ac04a207e69ba9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:58:30.923230Z","signature_b64":"rhc/WJIDvt7Bo03wZnh4oVS24E7PuVs08z0p9H7RbzHAh79DSxM4YLrffu15u3hTEmY1q1sMgWoQdQ7V1ZNXCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5c9ad4047220b197d789b3c8f632b2b9b13bda066f2f078c324d9e4c8a7b0709","last_reissued_at":"2026-07-05T09:58:30.922711Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:58:30.922711Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MM-GEN: Enhancing Task Performance Through Targeted Multimodal Data Curation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Baharan Mirzasoleiman, Besmira Nushi, Neel Joshi, Siddharth Joshi, Varun Chandrasekaran, Vibhav Vineet, Vidhisha Balachandran","submitted_at":"2025-01-07T21:55:56Z","abstract_excerpt":"Vision-language models (VLMs) are highly effective but often underperform on specialized tasks; for example, Llava-1.5 struggles with chart and diagram understanding due to scarce task-specific training data. Existing training data, sourced from general-purpose datasets, fails to capture the nuanced details needed for these tasks. We introduce MM-Gen, a scalable method that generates task-specific, high-quality synthetic text for candidate images by leveraging stronger models. MM-Gen employs a three-stage targeted process: partitioning data into subgroups, generating targeted text based on tas"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.04155","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/2501.04155/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":"2501.04155","created_at":"2026-07-05T09:58:30.922762+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.04155v1","created_at":"2026-07-05T09:58:30.922762+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.04155","created_at":"2026-07-05T09:58:30.922762+00:00"},{"alias_kind":"pith_short_12","alias_value":"LSNNIBDSECYZ","created_at":"2026-07-05T09:58:30.922762+00:00"},{"alias_kind":"pith_short_16","alias_value":"LSNNIBDSECYZPV4J","created_at":"2026-07-05T09:58:30.922762+00:00"},{"alias_kind":"pith_short_8","alias_value":"LSNNIBDS","created_at":"2026-07-05T09:58:30.922762+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.11405","citing_title":"20/20 Vision Language Models: A Prescription for Better VLMs through Data Curation Alone","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11405","citing_title":"20/20 Vision Language Models: A Prescription for Better VLMs through Data Curation Alone","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11634","citing_title":"Unlocking UML Class Diagram Understanding in Vision Language Models","ref_index":20,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LSNNIBDSECYZPV4JWPEPMMVSXG","json":"https://pith.science/pith/LSNNIBDSECYZPV4JWPEPMMVSXG.json","graph_json":"https://pith.science/api/pith-number/LSNNIBDSECYZPV4JWPEPMMVSXG/graph.json","events_json":"https://pith.science/api/pith-number/LSNNIBDSECYZPV4JWPEPMMVSXG/events.json","paper":"https://pith.science/paper/LSNNIBDS"},"agent_actions":{"view_html":"https://pith.science/pith/LSNNIBDSECYZPV4JWPEPMMVSXG","download_json":"https://pith.science/pith/LSNNIBDSECYZPV4JWPEPMMVSXG.json","view_paper":"https://pith.science/paper/LSNNIBDS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.04155&json=true","fetch_graph":"https://pith.science/api/pith-number/LSNNIBDSECYZPV4JWPEPMMVSXG/graph.json","fetch_events":"https://pith.science/api/pith-number/LSNNIBDSECYZPV4JWPEPMMVSXG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LSNNIBDSECYZPV4JWPEPMMVSXG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LSNNIBDSECYZPV4JWPEPMMVSXG/action/storage_attestation","attest_author":"https://pith.science/pith/LSNNIBDSECYZPV4JWPEPMMVSXG/action/author_attestation","sign_citation":"https://pith.science/pith/LSNNIBDSECYZPV4JWPEPMMVSXG/action/citation_signature","submit_replication":"https://pith.science/pith/LSNNIBDSECYZPV4JWPEPMMVSXG/action/replication_record"}},"created_at":"2026-07-05T09:58:30.922762+00:00","updated_at":"2026-07-05T09:58:30.922762+00:00"}