{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:IBR6CCF2HY2S5DUHYHSDT445CW","short_pith_number":"pith:IBR6CCF2","schema_version":"1.0","canonical_sha256":"4063e108ba3e352e8e87c1e439f39d15abf35b53adcae674e58b0cb9eed4f54b","source":{"kind":"arxiv","id":"2505.00703","version":2},"attestation_state":"computed","paper":{"title":"T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Dongzhi Jiang, Hao Li, Hongsheng Li, Le Zhuo, Pheng-Ann Heng, Renrui Zhang, Shilin Yan, Zhuofan Zong, Ziyu Guo","submitted_at":"2025-05-01T17:59:46Z","abstract_excerpt":"Recent advancements in large language models have demonstrated how chain-of-thought (CoT) and reinforcement learning (RL) can improve performance. However, applying such reasoning strategies to the visual generation domain remains largely unexplored. In this paper, we present T2I-R1, a novel reasoning-enhanced text-to-image generation model, powered by RL with a bi-level CoT reasoning process. Specifically, we identify two levels of CoT that can be utilized to enhance different stages of generation: (1) the semantic-level CoT for high-level planning of the prompt and (2) the token-level CoT fo"},"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":"2505.00703","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-05-01T17:59:46Z","cross_cats_sorted":["cs.AI","cs.CL","cs.LG"],"title_canon_sha256":"2093a6e0eb32cd6b7f1e5a33f8efc29f863f550b5cce98c06d3ce655ffc341c6","abstract_canon_sha256":"fad20bcb67c23761ef082454baa64e70f5d77663d34a27b4537750705c0d8cc5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:29:58.935869Z","signature_b64":"POOzzC4jGHkl7MypfJY5IhRxWdLFtvhrZ5Jg02Eby8b4rQuT2b8pKXTpblu2667KzI/QTe9oEMbUkbawjeacAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4063e108ba3e352e8e87c1e439f39d15abf35b53adcae674e58b0cb9eed4f54b","last_reissued_at":"2026-07-05T11:29:58.935375Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:29:58.935375Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Dongzhi Jiang, Hao Li, Hongsheng Li, Le Zhuo, Pheng-Ann Heng, Renrui Zhang, Shilin Yan, Zhuofan Zong, Ziyu Guo","submitted_at":"2025-05-01T17:59:46Z","abstract_excerpt":"Recent advancements in large language models have demonstrated how chain-of-thought (CoT) and reinforcement learning (RL) can improve performance. However, applying such reasoning strategies to the visual generation domain remains largely unexplored. In this paper, we present T2I-R1, a novel reasoning-enhanced text-to-image generation model, powered by RL with a bi-level CoT reasoning process. Specifically, we identify two levels of CoT that can be utilized to enhance different stages of generation: (1) the semantic-level CoT for high-level planning of the prompt and (2) the token-level CoT fo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.00703","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/2505.00703/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":"2505.00703","created_at":"2026-07-05T11:29:58.935437+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.00703v2","created_at":"2026-07-05T11:29:58.935437+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.00703","created_at":"2026-07-05T11:29:58.935437+00:00"},{"alias_kind":"pith_short_12","alias_value":"IBR6CCF2HY2S","created_at":"2026-07-05T11:29:58.935437+00:00"},{"alias_kind":"pith_short_16","alias_value":"IBR6CCF2HY2S5DUH","created_at":"2026-07-05T11:29:58.935437+00:00"},{"alias_kind":"pith_short_8","alias_value":"IBR6CCF2","created_at":"2026-07-05T11:29:58.935437+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":27,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26930","citing_title":"PortraitGen: Exemplar-Driven GRPO with Dual-Reward Guidance for Photorealistic Portrait Generation","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.17888","citing_title":"MathVis-Fine: Aligning Visual Supervision with Necessity via Progressive Dependency-Guided Training for Multimodal Mathematical Reasoning","ref_index":102,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05031","citing_title":"MetaPoint: Unlocking Precise Spatial Control in Agentic Visual Generation","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04264","citing_title":"UniCanvas: A Diffusion-base Unified Model for Text-in-Image Joint Generation","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31711","citing_title":"Arena-T2I Hard: Benchmarking and Improving Faithfulness with Dependency-Aware Checklist","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25343","citing_title":"Toward Native Multimodal Modeling: A Roadmap","ref_index":191,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28615","citing_title":"Compositional Text-to-Image Generation Via Region-aware Bimodal Direct Preference Optimization","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30248","citing_title":"GenClaw: Code-Driven Agentic Image Generation","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17602","citing_title":"AutoRubric-T2I: Robust Rule-Based Reward Model for Text-to-Image Alignment","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22564","citing_title":"SynAE: A Framework for Measuring the Quality of Synthetic Data for Tool-Calling Agent Evaluations","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21195","citing_title":"RankE: End-to-End Post-Training for Discrete Text-to-Image Generation with Decoder Co-Evolution","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17602","citing_title":"AutoRubric-T2I: Robust Rule-Based Reward Model for Text-to-Image Alignment","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14876","citing_title":"Unlocking Complex Visual Generation via Closed-Loop Verified Reasoning","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2506.18871","citing_title":"OmniGen2: Towards Instruction-Aligned Multimodal Generation","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2509.02547","citing_title":"The Landscape of Agentic Reinforcement Learning for LLMs: A Survey","ref_index":259,"is_internal_anchor":false},{"citing_arxiv_id":"2509.08827","citing_title":"A Survey of Reinforcement Learning for Large Reasoning Models","ref_index":238,"is_internal_anchor":false},{"citing_arxiv_id":"2512.01236","citing_title":"PSR: Scaling Multi-Subject Personalized Image Generation with Pairwise Subject-Consistency Rewards","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2512.07348","citing_title":"MICo-150K: A Comprehensive Dataset Advancing Multi-Image Composition","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2602.19974","citing_title":"RL-RIG: A Generative Spatial Reasoner via Intrinsic Reflection","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02355","citing_title":"From Broad Exploration to Stable Synthesis: Entropy-Guided Optimization for Autoregressive Image Generation","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12480","citing_title":"OmniNFT: Modality-wise Omni Diffusion Reinforcement for Joint Audio-Video Generation","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11400","citing_title":"UniPath: Adaptive Coordination of Understanding and Generation for Unified Multimodal Reasoning","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24625","citing_title":"Meta-CoT: Enhancing Granularity and Generalization in Image Editing","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12617","citing_title":"SOAR: Self-Correction for Optimal Alignment and Refinement in Diffusion Models","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06966","citing_title":"MAR-GRPO: Stabilized GRPO for AR-diffusion Hybrid Image Generation","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IBR6CCF2HY2S5DUHYHSDT445CW","json":"https://pith.science/pith/IBR6CCF2HY2S5DUHYHSDT445CW.json","graph_json":"https://pith.science/api/pith-number/IBR6CCF2HY2S5DUHYHSDT445CW/graph.json","events_json":"https://pith.science/api/pith-number/IBR6CCF2HY2S5DUHYHSDT445CW/events.json","paper":"https://pith.science/paper/IBR6CCF2"},"agent_actions":{"view_html":"https://pith.science/pith/IBR6CCF2HY2S5DUHYHSDT445CW","download_json":"https://pith.science/pith/IBR6CCF2HY2S5DUHYHSDT445CW.json","view_paper":"https://pith.science/paper/IBR6CCF2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.00703&json=true","fetch_graph":"https://pith.science/api/pith-number/IBR6CCF2HY2S5DUHYHSDT445CW/graph.json","fetch_events":"https://pith.science/api/pith-number/IBR6CCF2HY2S5DUHYHSDT445CW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IBR6CCF2HY2S5DUHYHSDT445CW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IBR6CCF2HY2S5DUHYHSDT445CW/action/storage_attestation","attest_author":"https://pith.science/pith/IBR6CCF2HY2S5DUHYHSDT445CW/action/author_attestation","sign_citation":"https://pith.science/pith/IBR6CCF2HY2S5DUHYHSDT445CW/action/citation_signature","submit_replication":"https://pith.science/pith/IBR6CCF2HY2S5DUHYHSDT445CW/action/replication_record"}},"created_at":"2026-07-05T11:29:58.935437+00:00","updated_at":"2026-07-05T11:29:58.935437+00:00"}