{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:4RB5CNWTI5C4PBQQBBKYVMGAIC","short_pith_number":"pith:4RB5CNWT","schema_version":"1.0","canonical_sha256":"e443d136d34745c7861008558ab0c04086857597a937daef015ee5b37218ed3f","source":{"kind":"arxiv","id":"2408.13256","version":2},"attestation_state":"computed","paper":{"title":"How Diffusion Models Learn to Factorize and Compose","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.AI","authors_text":"Ila Fiete, Mitchell Ostrow, Qiyao Liang, Ziming Liu","submitted_at":"2024-08-23T17:59:03Z","abstract_excerpt":"Diffusion models are capable of generating photo-realistic images that combine elements which likely do not appear together in the training set, demonstrating the ability to \\textit{compositionally generalize}. Nonetheless, the precise mechanism of compositionality and how it is acquired through training remains elusive. Inspired by cognitive neuroscientific approaches, we consider a highly reduced setting to examine whether and when diffusion models learn semantically meaningful and factorized representations of composable features. We performed extensive controlled experiments on conditional"},"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":"2408.13256","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.AI","submitted_at":"2024-08-23T17:59:03Z","cross_cats_sorted":["cs.CV","cs.LG"],"title_canon_sha256":"cae5d648650a218ce5055f56b5364c873ffbcfb085a477a10a9d9d5f0f7a38b2","abstract_canon_sha256":"d0e23c22a614297f79b9ce03e01b6d54d70e98c1f465914635b07181bb61b08e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:18:53.776401Z","signature_b64":"MYujLaFRCaXSWTEyZGslMjVbtDR2fOnNLz1WV4/J+HdaUAQQuWyYa00zEq+AR2BgjZ3hCvrS0ltGRpJ89Ug1Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e443d136d34745c7861008558ab0c04086857597a937daef015ee5b37218ed3f","last_reissued_at":"2026-07-05T09:18:53.775912Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:18:53.775912Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Diffusion Models Learn to Factorize and Compose","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.AI","authors_text":"Ila Fiete, Mitchell Ostrow, Qiyao Liang, Ziming Liu","submitted_at":"2024-08-23T17:59:03Z","abstract_excerpt":"Diffusion models are capable of generating photo-realistic images that combine elements which likely do not appear together in the training set, demonstrating the ability to \\textit{compositionally generalize}. Nonetheless, the precise mechanism of compositionality and how it is acquired through training remains elusive. Inspired by cognitive neuroscientific approaches, we consider a highly reduced setting to examine whether and when diffusion models learn semantically meaningful and factorized representations of composable features. We performed extensive controlled experiments on conditional"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.13256","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/2408.13256/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":"2408.13256","created_at":"2026-07-05T09:18:53.775974+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.13256v2","created_at":"2026-07-05T09:18:53.775974+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.13256","created_at":"2026-07-05T09:18:53.775974+00:00"},{"alias_kind":"pith_short_12","alias_value":"4RB5CNWTI5C4","created_at":"2026-07-05T09:18:53.775974+00:00"},{"alias_kind":"pith_short_16","alias_value":"4RB5CNWTI5C4PBQQ","created_at":"2026-07-05T09:18:53.775974+00:00"},{"alias_kind":"pith_short_8","alias_value":"4RB5CNWT","created_at":"2026-07-05T09:18:53.775974+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.27696","citing_title":"Structure over Pixels: Learning Variable-Length Visual Programs","ref_index":38,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4RB5CNWTI5C4PBQQBBKYVMGAIC","json":"https://pith.science/pith/4RB5CNWTI5C4PBQQBBKYVMGAIC.json","graph_json":"https://pith.science/api/pith-number/4RB5CNWTI5C4PBQQBBKYVMGAIC/graph.json","events_json":"https://pith.science/api/pith-number/4RB5CNWTI5C4PBQQBBKYVMGAIC/events.json","paper":"https://pith.science/paper/4RB5CNWT"},"agent_actions":{"view_html":"https://pith.science/pith/4RB5CNWTI5C4PBQQBBKYVMGAIC","download_json":"https://pith.science/pith/4RB5CNWTI5C4PBQQBBKYVMGAIC.json","view_paper":"https://pith.science/paper/4RB5CNWT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.13256&json=true","fetch_graph":"https://pith.science/api/pith-number/4RB5CNWTI5C4PBQQBBKYVMGAIC/graph.json","fetch_events":"https://pith.science/api/pith-number/4RB5CNWTI5C4PBQQBBKYVMGAIC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4RB5CNWTI5C4PBQQBBKYVMGAIC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4RB5CNWTI5C4PBQQBBKYVMGAIC/action/storage_attestation","attest_author":"https://pith.science/pith/4RB5CNWTI5C4PBQQBBKYVMGAIC/action/author_attestation","sign_citation":"https://pith.science/pith/4RB5CNWTI5C4PBQQBBKYVMGAIC/action/citation_signature","submit_replication":"https://pith.science/pith/4RB5CNWTI5C4PBQQBBKYVMGAIC/action/replication_record"}},"created_at":"2026-07-05T09:18:53.775974+00:00","updated_at":"2026-07-05T09:18:53.775974+00:00"}