{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:7VCN6UPUP37CJP6EV6TJYSV2TL","short_pith_number":"pith:7VCN6UPU","schema_version":"1.0","canonical_sha256":"fd44df51f47efe24bfc4afa69c4aba9af25070ac41d2f9a52980155398ee26c9","source":{"kind":"arxiv","id":"2511.20651","version":2},"attestation_state":"computed","paper":{"title":"RubricRL: Simple Generalizable Rewards for Text-to-Image Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chunming Qiao, Dongdong Chen, Junsong Yuan, Xuelu Feng, Yunsheng Li, Zixuan Gao, Ziyu Wan","submitted_at":"2025-11-25T18:59:55Z","abstract_excerpt":"Reinforcement learning (RL) has recently emerged as a promising approach for aligning text-to-image generative models with human preferences. A key challenge, however, lies in designing effective and interpretable rewards. Existing methods often rely on either composite metrics (e.g., CLIP, OCR, and realism scores) with fixed weights or a single scalar reward distilled from human preference models, which can limit interpretability and flexibility. We propose RubricRL, a simple and general framework for rubric-based reward design that offers greater interpretability, composability, and user con"},"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":"2511.20651","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-11-25T18:59:55Z","cross_cats_sorted":[],"title_canon_sha256":"ab53405f6b98f444ac18d43eb3ad5346eaf42c7907313f67e535f537b51b9afd","abstract_canon_sha256":"9fc8b8a93ef61713949c8531576181b99c7d97af3bd48297203ad0fe31878ca1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-23T03:14:27.684078Z","signature_b64":"9b9WKDT5qcmO0dT8fpAg8a7jDTnz6zZ9PTCXt8iwv7jQ4rIK7jrqwHiXopjRCxDE6Rqf2NkLWH3RgB+Xo6myAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fd44df51f47efe24bfc4afa69c4aba9af25070ac41d2f9a52980155398ee26c9","last_reissued_at":"2026-06-23T03:14:27.683579Z","signature_status":"signed_v1","first_computed_at":"2026-06-23T03:14:27.683579Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RubricRL: Simple Generalizable Rewards for Text-to-Image Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chunming Qiao, Dongdong Chen, Junsong Yuan, Xuelu Feng, Yunsheng Li, Zixuan Gao, Ziyu Wan","submitted_at":"2025-11-25T18:59:55Z","abstract_excerpt":"Reinforcement learning (RL) has recently emerged as a promising approach for aligning text-to-image generative models with human preferences. A key challenge, however, lies in designing effective and interpretable rewards. Existing methods often rely on either composite metrics (e.g., CLIP, OCR, and realism scores) with fixed weights or a single scalar reward distilled from human preference models, which can limit interpretability and flexibility. We propose RubricRL, a simple and general framework for rubric-based reward design that offers greater interpretability, composability, and user con"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2511.20651","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/2511.20651/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":"2511.20651","created_at":"2026-06-23T03:14:27.683641+00:00"},{"alias_kind":"arxiv_version","alias_value":"2511.20651v2","created_at":"2026-06-23T03:14:27.683641+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.20651","created_at":"2026-06-23T03:14:27.683641+00:00"},{"alias_kind":"pith_short_12","alias_value":"7VCN6UPUP37C","created_at":"2026-06-23T03:14:27.683641+00:00"},{"alias_kind":"pith_short_16","alias_value":"7VCN6UPUP37CJP6E","created_at":"2026-06-23T03:14:27.683641+00:00"},{"alias_kind":"pith_short_8","alias_value":"7VCN6UPU","created_at":"2026-06-23T03:14:27.683641+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"2605.21573","citing_title":"Lens: Rethinking Training Efficiency for Foundational Text-to-Image Models","ref_index":25,"is_internal_anchor":true},{"citing_arxiv_id":"2605.17602","citing_title":"AutoRubric-T2I: Robust Rule-Based Reward Model for Text-to-Image Alignment","ref_index":4,"is_internal_anchor":true},{"citing_arxiv_id":"2605.17602","citing_title":"AutoRubric-T2I: Robust Rule-Based Reward Model for Text-to-Image Alignment","ref_index":4,"is_internal_anchor":true},{"citing_arxiv_id":"2604.13618","citing_title":"C2: Scalable Rubric-Augmented Reward Modeling from Binary Preferences","ref_index":3,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7VCN6UPUP37CJP6EV6TJYSV2TL","json":"https://pith.science/pith/7VCN6UPUP37CJP6EV6TJYSV2TL.json","graph_json":"https://pith.science/api/pith-number/7VCN6UPUP37CJP6EV6TJYSV2TL/graph.json","events_json":"https://pith.science/api/pith-number/7VCN6UPUP37CJP6EV6TJYSV2TL/events.json","paper":"https://pith.science/paper/7VCN6UPU"},"agent_actions":{"view_html":"https://pith.science/pith/7VCN6UPUP37CJP6EV6TJYSV2TL","download_json":"https://pith.science/pith/7VCN6UPUP37CJP6EV6TJYSV2TL.json","view_paper":"https://pith.science/paper/7VCN6UPU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2511.20651&json=true","fetch_graph":"https://pith.science/api/pith-number/7VCN6UPUP37CJP6EV6TJYSV2TL/graph.json","fetch_events":"https://pith.science/api/pith-number/7VCN6UPUP37CJP6EV6TJYSV2TL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7VCN6UPUP37CJP6EV6TJYSV2TL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7VCN6UPUP37CJP6EV6TJYSV2TL/action/storage_attestation","attest_author":"https://pith.science/pith/7VCN6UPUP37CJP6EV6TJYSV2TL/action/author_attestation","sign_citation":"https://pith.science/pith/7VCN6UPUP37CJP6EV6TJYSV2TL/action/citation_signature","submit_replication":"https://pith.science/pith/7VCN6UPUP37CJP6EV6TJYSV2TL/action/replication_record"}},"created_at":"2026-06-23T03:14:27.683641+00:00","updated_at":"2026-06-23T03:14:27.683641+00:00"}