{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:HT52AJRVGYWEC5XM2EGK6V7G3U","short_pith_number":"pith:HT52AJRV","schema_version":"1.0","canonical_sha256":"3cfba02635362c4176ecd10caf57e6dd1b57235a7b917dff2b52a84cbac015c1","source":{"kind":"arxiv","id":"2503.13436","version":1},"attestation_state":"computed","paper":{"title":"Unified Autoregressive Visual Generation and Understanding with Continuous Tokens","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Andreas Steiner, Chen Sun, Deqing Sun, Lijie Fan, Luming Tang, Michael Rubinstein, Michalis Raptis, Radu Soricut, Siyang Qin, Siyuan Qiao, Tao Zhu, Tianhong Li, Xuan Yang, Yuanzhen Li","submitted_at":"2025-03-17T17:58:30Z","abstract_excerpt":"We present UniFluid, a unified autoregressive framework for joint visual generation and understanding leveraging continuous visual tokens. Our unified autoregressive architecture processes multimodal image and text inputs, generating discrete tokens for text and continuous tokens for image. We find though there is an inherent trade-off between the image generation and understanding task, a carefully tuned training recipe enables them to improve each other. By selecting an appropriate loss balance weight, the unified model achieves results comparable to or exceeding those of single-task baselin"},"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":"2503.13436","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-17T17:58:30Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"ca77c173374d9ed8b4a167c86937f5e4f3eaaa6c6a20150d9a5fea079f9851c4","abstract_canon_sha256":"254624996901940401c0d5111fe440031961a2cc505e11ecbc40da9c40774619"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:33:00.678628Z","signature_b64":"DbH7WpOEzt3mIAB/atyxCrNOmAofI9aY0ubzW3zq0jHEoOHTi1FHgGF60azSZwdZS9rfy/KBSxW7J4S8A98NBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3cfba02635362c4176ecd10caf57e6dd1b57235a7b917dff2b52a84cbac015c1","last_reissued_at":"2026-07-05T10:33:00.678131Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:33:00.678131Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unified Autoregressive Visual Generation and Understanding with Continuous Tokens","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Andreas Steiner, Chen Sun, Deqing Sun, Lijie Fan, Luming Tang, Michael Rubinstein, Michalis Raptis, Radu Soricut, Siyang Qin, Siyuan Qiao, Tao Zhu, Tianhong Li, Xuan Yang, Yuanzhen Li","submitted_at":"2025-03-17T17:58:30Z","abstract_excerpt":"We present UniFluid, a unified autoregressive framework for joint visual generation and understanding leveraging continuous visual tokens. Our unified autoregressive architecture processes multimodal image and text inputs, generating discrete tokens for text and continuous tokens for image. We find though there is an inherent trade-off between the image generation and understanding task, a carefully tuned training recipe enables them to improve each other. By selecting an appropriate loss balance weight, the unified model achieves results comparable to or exceeding those of single-task baselin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.13436","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/2503.13436/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":"2503.13436","created_at":"2026-07-05T10:33:00.678190+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.13436v1","created_at":"2026-07-05T10:33:00.678190+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.13436","created_at":"2026-07-05T10:33:00.678190+00:00"},{"alias_kind":"pith_short_12","alias_value":"HT52AJRVGYWE","created_at":"2026-07-05T10:33:00.678190+00:00"},{"alias_kind":"pith_short_16","alias_value":"HT52AJRVGYWEC5XM","created_at":"2026-07-05T10:33:00.678190+00:00"},{"alias_kind":"pith_short_8","alias_value":"HT52AJRV","created_at":"2026-07-05T10:33:00.678190+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.13289","citing_title":"HYDRA-X: Native Unified Multimodal Models with Holistic Visual Tokenizers","ref_index":280,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18678","citing_title":"Lance: Unified Multimodal Modeling by Multi-Task Synergy","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24763","citing_title":"Tuna-2: Pixel Embeddings Beat Vision Encoders for Multimodal Understanding and Generation","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18678","citing_title":"Lance: Unified Multimodal Modeling by Multi-Task Synergy","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2512.13592","citing_title":"Image Diffusion Preview with Consistency Solver","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03893","citing_title":"FeynmanBench: Benchmarking Multimodal LLMs on Diagrammatic Physics Reasoning","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2506.15564","citing_title":"Show-o2: Improved Native Unified Multimodal Models","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24763","citing_title":"Tuna-2: Pixel Embeddings Beat Vision Encoders for Multimodal Understanding and Generation","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2504.17761","citing_title":"Step1X-Edit: A Practical Framework for General Image Editing","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10949","citing_title":"Pseudo-Unification: Entropy Probing Reveals Divergent Information Patterns in Unified Multimodal Models","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HT52AJRVGYWEC5XM2EGK6V7G3U","json":"https://pith.science/pith/HT52AJRVGYWEC5XM2EGK6V7G3U.json","graph_json":"https://pith.science/api/pith-number/HT52AJRVGYWEC5XM2EGK6V7G3U/graph.json","events_json":"https://pith.science/api/pith-number/HT52AJRVGYWEC5XM2EGK6V7G3U/events.json","paper":"https://pith.science/paper/HT52AJRV"},"agent_actions":{"view_html":"https://pith.science/pith/HT52AJRVGYWEC5XM2EGK6V7G3U","download_json":"https://pith.science/pith/HT52AJRVGYWEC5XM2EGK6V7G3U.json","view_paper":"https://pith.science/paper/HT52AJRV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.13436&json=true","fetch_graph":"https://pith.science/api/pith-number/HT52AJRVGYWEC5XM2EGK6V7G3U/graph.json","fetch_events":"https://pith.science/api/pith-number/HT52AJRVGYWEC5XM2EGK6V7G3U/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HT52AJRVGYWEC5XM2EGK6V7G3U/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HT52AJRVGYWEC5XM2EGK6V7G3U/action/storage_attestation","attest_author":"https://pith.science/pith/HT52AJRVGYWEC5XM2EGK6V7G3U/action/author_attestation","sign_citation":"https://pith.science/pith/HT52AJRVGYWEC5XM2EGK6V7G3U/action/citation_signature","submit_replication":"https://pith.science/pith/HT52AJRVGYWEC5XM2EGK6V7G3U/action/replication_record"}},"created_at":"2026-07-05T10:33:00.678190+00:00","updated_at":"2026-07-05T10:33:00.678190+00:00"}