{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ULW6Q5O45IB6DJTL57BOUG3JIO","short_pith_number":"pith:ULW6Q5O4","schema_version":"1.0","canonical_sha256":"a2ede875dcea03e1a66befc2ea1b6943b066c198a1e43c87d921bb5fda4a35eb","source":{"kind":"arxiv","id":"2505.14682","version":1},"attestation_state":"computed","paper":{"title":"UniGen: Enhanced Training & Test-Time Strategies for Unified Multimodal Understanding and Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Afshin Dehghan, Jiaming Hu, Jiasen Lu, Mingfei Gao, Mingze Xu, Rui Tian, Yinfei Yang, Zuxuan Wu","submitted_at":"2025-05-20T17:59:26Z","abstract_excerpt":"We introduce UniGen, a unified multimodal large language model (MLLM) capable of image understanding and generation. We study the full training pipeline of UniGen from a data-centric perspective, including multi-stage pre-training, supervised fine-tuning, and direct preference optimization. More importantly, we propose a new Chain-of-Thought Verification (CoT-V) strategy for test-time scaling, which significantly boosts UniGen's image generation quality using a simple Best-of-N test-time strategy. Specifically, CoT-V enables UniGen to act as both image generator and verifier at test time, asse"},"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":"2505.14682","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-05-20T17:59:26Z","cross_cats_sorted":[],"title_canon_sha256":"7b5bf3ccedd99ac8df2b22c78718d01cad6664f422564dcdf6a8bcf92d414ff9","abstract_canon_sha256":"d1436f589d43b92cc6f929562625ab4ade4ef17ce8c96a55b183d3c728c1dff5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:06:17.619683Z","signature_b64":"uex8qRX09pC2lNYgnGrdygNP7ucnjK0emPOGbgDHPkV5Uoi/0SZAkCte3odtkehdhhUrZuji/DCds32e2m5HCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a2ede875dcea03e1a66befc2ea1b6943b066c198a1e43c87d921bb5fda4a35eb","last_reissued_at":"2026-07-05T11:06:17.619109Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:06:17.619109Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"UniGen: Enhanced Training & Test-Time Strategies for Unified Multimodal Understanding and Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Afshin Dehghan, Jiaming Hu, Jiasen Lu, Mingfei Gao, Mingze Xu, Rui Tian, Yinfei Yang, Zuxuan Wu","submitted_at":"2025-05-20T17:59:26Z","abstract_excerpt":"We introduce UniGen, a unified multimodal large language model (MLLM) capable of image understanding and generation. We study the full training pipeline of UniGen from a data-centric perspective, including multi-stage pre-training, supervised fine-tuning, and direct preference optimization. More importantly, we propose a new Chain-of-Thought Verification (CoT-V) strategy for test-time scaling, which significantly boosts UniGen's image generation quality using a simple Best-of-N test-time strategy. Specifically, CoT-V enables UniGen to act as both image generator and verifier at test time, asse"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.14682","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/2505.14682/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":"2505.14682","created_at":"2026-07-05T11:06:17.619176+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.14682v1","created_at":"2026-07-05T11:06:17.619176+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.14682","created_at":"2026-07-05T11:06:17.619176+00:00"},{"alias_kind":"pith_short_12","alias_value":"ULW6Q5O45IB6","created_at":"2026-07-05T11:06:17.619176+00:00"},{"alias_kind":"pith_short_16","alias_value":"ULW6Q5O45IB6DJTL","created_at":"2026-07-05T11:06:17.619176+00:00"},{"alias_kind":"pith_short_8","alias_value":"ULW6Q5O4","created_at":"2026-07-05T11:06:17.619176+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.08231","citing_title":"Test-Time Scaling in Multimodal Foundation Models: A Comprehensive Survey of Generation and Reasoning","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2506.15564","citing_title":"Show-o2: Improved Native Unified Multimodal Models","ref_index":104,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ULW6Q5O45IB6DJTL57BOUG3JIO","json":"https://pith.science/pith/ULW6Q5O45IB6DJTL57BOUG3JIO.json","graph_json":"https://pith.science/api/pith-number/ULW6Q5O45IB6DJTL57BOUG3JIO/graph.json","events_json":"https://pith.science/api/pith-number/ULW6Q5O45IB6DJTL57BOUG3JIO/events.json","paper":"https://pith.science/paper/ULW6Q5O4"},"agent_actions":{"view_html":"https://pith.science/pith/ULW6Q5O45IB6DJTL57BOUG3JIO","download_json":"https://pith.science/pith/ULW6Q5O45IB6DJTL57BOUG3JIO.json","view_paper":"https://pith.science/paper/ULW6Q5O4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.14682&json=true","fetch_graph":"https://pith.science/api/pith-number/ULW6Q5O45IB6DJTL57BOUG3JIO/graph.json","fetch_events":"https://pith.science/api/pith-number/ULW6Q5O45IB6DJTL57BOUG3JIO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ULW6Q5O45IB6DJTL57BOUG3JIO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ULW6Q5O45IB6DJTL57BOUG3JIO/action/storage_attestation","attest_author":"https://pith.science/pith/ULW6Q5O45IB6DJTL57BOUG3JIO/action/author_attestation","sign_citation":"https://pith.science/pith/ULW6Q5O45IB6DJTL57BOUG3JIO/action/citation_signature","submit_replication":"https://pith.science/pith/ULW6Q5O45IB6DJTL57BOUG3JIO/action/replication_record"}},"created_at":"2026-07-05T11:06:17.619176+00:00","updated_at":"2026-07-05T11:06:17.619176+00:00"}