{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:72SCPPF3CKZRJRQIOAVLPQGZ2I","short_pith_number":"pith:72SCPPF3","schema_version":"1.0","canonical_sha256":"fea427bcbb12b314c608702ab7c0d9d22649bd1ddd31945b2c8c5306942f1f38","source":{"kind":"arxiv","id":"2505.11178","version":1},"attestation_state":"computed","paper":{"title":"CompAlign: Improving Compositional Text-to-Image Generation with a Complex Benchmark and Fine-Grained Feedback","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CV","authors_text":"Kai-Wei Chang, Yixin Wan","submitted_at":"2025-05-16T12:23:58Z","abstract_excerpt":"State-of-the-art T2I models are capable of generating high-resolution images given textual prompts. However, they still struggle with accurately depicting compositional scenes that specify multiple objects, attributes, and spatial relations. We present CompAlign, a challenging benchmark with an emphasis on assessing the depiction of 3D-spatial relationships, for evaluating and improving models on compositional image generation. CompAlign consists of 900 complex multi-subject image generation prompts that combine numerical and 3D-spatial relationships with varied attribute bindings. Our benchma"},"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":"2505.11178","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.CV","submitted_at":"2025-05-16T12:23:58Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"5cc9738b10e9594625653099d289d4e6cfb56a2dfa9b0ff525e726bb5c5961fc","abstract_canon_sha256":"f54a41c5b2f8dc9ad34b8b0afd10ea777d39b401740db574ee25f54277df3699"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:04:12.149596Z","signature_b64":"8VupvurrrlIXeiXxEqxsvovsvoc+h1CYa1WvL8JdWos0yDqSM9JLEhG9ujKj/DUQo2NbxVv7q9yBl6V441gMBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fea427bcbb12b314c608702ab7c0d9d22649bd1ddd31945b2c8c5306942f1f38","last_reissued_at":"2026-07-05T11:04:12.149139Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:04:12.149139Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CompAlign: Improving Compositional Text-to-Image Generation with a Complex Benchmark and Fine-Grained Feedback","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CV","authors_text":"Kai-Wei Chang, Yixin Wan","submitted_at":"2025-05-16T12:23:58Z","abstract_excerpt":"State-of-the-art T2I models are capable of generating high-resolution images given textual prompts. However, they still struggle with accurately depicting compositional scenes that specify multiple objects, attributes, and spatial relations. We present CompAlign, a challenging benchmark with an emphasis on assessing the depiction of 3D-spatial relationships, for evaluating and improving models on compositional image generation. CompAlign consists of 900 complex multi-subject image generation prompts that combine numerical and 3D-spatial relationships with varied attribute bindings. Our benchma"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.11178","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/2505.11178/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":"2505.11178","created_at":"2026-07-05T11:04:12.149204+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.11178v1","created_at":"2026-07-05T11:04:12.149204+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.11178","created_at":"2026-07-05T11:04:12.149204+00:00"},{"alias_kind":"pith_short_12","alias_value":"72SCPPF3CKZR","created_at":"2026-07-05T11:04:12.149204+00:00"},{"alias_kind":"pith_short_16","alias_value":"72SCPPF3CKZRJRQI","created_at":"2026-07-05T11:04:12.149204+00:00"},{"alias_kind":"pith_short_8","alias_value":"72SCPPF3","created_at":"2026-07-05T11:04:12.149204+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24849","citing_title":"IV-CoT: Implicit Visual Chain-of-Thought for Structure-Aware Text-to-Image Generation","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06170","citing_title":"DynT2I-Eval: A Dynamic Evaluation Framework for Text-to-Image Models","ref_index":36,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/72SCPPF3CKZRJRQIOAVLPQGZ2I","json":"https://pith.science/pith/72SCPPF3CKZRJRQIOAVLPQGZ2I.json","graph_json":"https://pith.science/api/pith-number/72SCPPF3CKZRJRQIOAVLPQGZ2I/graph.json","events_json":"https://pith.science/api/pith-number/72SCPPF3CKZRJRQIOAVLPQGZ2I/events.json","paper":"https://pith.science/paper/72SCPPF3"},"agent_actions":{"view_html":"https://pith.science/pith/72SCPPF3CKZRJRQIOAVLPQGZ2I","download_json":"https://pith.science/pith/72SCPPF3CKZRJRQIOAVLPQGZ2I.json","view_paper":"https://pith.science/paper/72SCPPF3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.11178&json=true","fetch_graph":"https://pith.science/api/pith-number/72SCPPF3CKZRJRQIOAVLPQGZ2I/graph.json","fetch_events":"https://pith.science/api/pith-number/72SCPPF3CKZRJRQIOAVLPQGZ2I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/72SCPPF3CKZRJRQIOAVLPQGZ2I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/72SCPPF3CKZRJRQIOAVLPQGZ2I/action/storage_attestation","attest_author":"https://pith.science/pith/72SCPPF3CKZRJRQIOAVLPQGZ2I/action/author_attestation","sign_citation":"https://pith.science/pith/72SCPPF3CKZRJRQIOAVLPQGZ2I/action/citation_signature","submit_replication":"https://pith.science/pith/72SCPPF3CKZRJRQIOAVLPQGZ2I/action/replication_record"}},"created_at":"2026-07-05T11:04:12.149204+00:00","updated_at":"2026-07-05T11:04:12.149204+00:00"}