{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RFI5KKZZZ7L3WJ4APACYF2Z4SY","short_pith_number":"pith:RFI5KKZZ","schema_version":"1.0","canonical_sha256":"8951d52b39cfd7bb2780780582eb3c9634d88aea3df6e70cbd5889f7b2fc9250","source":{"kind":"arxiv","id":"2603.02667","version":2},"attestation_state":"computed","paper":{"title":"Unifying Contrastive and Generative Objectives for Visual Understanding and Text-to-Image Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Aashu Singh, Chao Li, Dina Katabi, Hong-You Chen, Jianpeng Cheng, Jun Xiao, Sai Vidyaranya Nuthalapati, Satya Narayan Shukla, Shlok Kumar Mishra, Tianhong Li, Xiangjun Fan, Yonghuan Yang","submitted_at":"2026-03-03T06:54:19Z","abstract_excerpt":"Unifying text-image contrastive learning and text-to-image (T2I) generation in a single end-to-end model is challenging because the two objectives demand opposing masking regimes: contrastive alignment needs near-complete visible tokens, while masked generative modeling needs heavy corruption. We introduce DREAM, a unified framework that resolves this conflict through Masking Warmup, a schedule that shifts the center of the masking distribution over training, so low and high masking ratios coexist at every step. This co-exposure lets a single jointly-trained encoder serve both objectives. The "},"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":"2603.02667","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-03-03T06:54:19Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"2b22dbed0fc1c07722929ab60c9b784aa12c9652c30ca2126d99158e8ec5a772","abstract_canon_sha256":"385672dc0784ac59116781e12db510f4d6296ec473c05f621bd5d2be80ad53d3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:07.668409Z","signature_b64":"CZtUj1TNZcGTdG4ikeaHMHApZVpeV+VCn0yCi9tyg7lTjcIuxntPZnN1cGGNHTyouUw1uWw6O6m7GGVyxbvECQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8951d52b39cfd7bb2780780582eb3c9634d88aea3df6e70cbd5889f7b2fc9250","last_reissued_at":"2026-05-20T00:03:07.667566Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:07.667566Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unifying Contrastive and Generative Objectives for Visual Understanding and Text-to-Image Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Aashu Singh, Chao Li, Dina Katabi, Hong-You Chen, Jianpeng Cheng, Jun Xiao, Sai Vidyaranya Nuthalapati, Satya Narayan Shukla, Shlok Kumar Mishra, Tianhong Li, Xiangjun Fan, Yonghuan Yang","submitted_at":"2026-03-03T06:54:19Z","abstract_excerpt":"Unifying text-image contrastive learning and text-to-image (T2I) generation in a single end-to-end model is challenging because the two objectives demand opposing masking regimes: contrastive alignment needs near-complete visible tokens, while masked generative modeling needs heavy corruption. We introduce DREAM, a unified framework that resolves this conflict through Masking Warmup, a schedule that shifts the center of the masking distribution over training, so low and high masking ratios coexist at every step. This co-exposure lets a single jointly-trained encoder serve both objectives. The "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.02667","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/2603.02667/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":"2603.02667","created_at":"2026-05-20T00:03:07.667706+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.02667v2","created_at":"2026-05-20T00:03:07.667706+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.02667","created_at":"2026-05-20T00:03:07.667706+00:00"},{"alias_kind":"pith_short_12","alias_value":"RFI5KKZZZ7L3","created_at":"2026-05-20T00:03:07.667706+00:00"},{"alias_kind":"pith_short_16","alias_value":"RFI5KKZZZ7L3WJ4A","created_at":"2026-05-20T00:03:07.667706+00:00"},{"alias_kind":"pith_short_8","alias_value":"RFI5KKZZ","created_at":"2026-05-20T00:03:07.667706+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/RFI5KKZZZ7L3WJ4APACYF2Z4SY","json":"https://pith.science/pith/RFI5KKZZZ7L3WJ4APACYF2Z4SY.json","graph_json":"https://pith.science/api/pith-number/RFI5KKZZZ7L3WJ4APACYF2Z4SY/graph.json","events_json":"https://pith.science/api/pith-number/RFI5KKZZZ7L3WJ4APACYF2Z4SY/events.json","paper":"https://pith.science/paper/RFI5KKZZ"},"agent_actions":{"view_html":"https://pith.science/pith/RFI5KKZZZ7L3WJ4APACYF2Z4SY","download_json":"https://pith.science/pith/RFI5KKZZZ7L3WJ4APACYF2Z4SY.json","view_paper":"https://pith.science/paper/RFI5KKZZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.02667&json=true","fetch_graph":"https://pith.science/api/pith-number/RFI5KKZZZ7L3WJ4APACYF2Z4SY/graph.json","fetch_events":"https://pith.science/api/pith-number/RFI5KKZZZ7L3WJ4APACYF2Z4SY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RFI5KKZZZ7L3WJ4APACYF2Z4SY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RFI5KKZZZ7L3WJ4APACYF2Z4SY/action/storage_attestation","attest_author":"https://pith.science/pith/RFI5KKZZZ7L3WJ4APACYF2Z4SY/action/author_attestation","sign_citation":"https://pith.science/pith/RFI5KKZZZ7L3WJ4APACYF2Z4SY/action/citation_signature","submit_replication":"https://pith.science/pith/RFI5KKZZZ7L3WJ4APACYF2Z4SY/action/replication_record"}},"created_at":"2026-05-20T00:03:07.667706+00:00","updated_at":"2026-05-20T00:03:07.667706+00:00"}