{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:QZ54L2VWCUZ4UDMZW734S3RTBU","short_pith_number":"pith:QZ54L2VW","schema_version":"1.0","canonical_sha256":"867bc5eab61533ca0d99b7f7c96e330d39ff6563d2b6662d82f3ca0ef1333440","source":{"kind":"arxiv","id":"2502.00968","version":2},"attestation_state":"computed","paper":{"title":"CoDe: Blockwise Control for Denoising Diffusion Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Ahmad Beirami, Anuj Singh, Hadi Jamali-Rad, Sayak Mukherjee","submitted_at":"2025-02-03T00:23:04Z","abstract_excerpt":"Aligning diffusion models to downstream tasks often requires finetuning new models or gradient-based guidance at inference time to enable sampling from the reward-tilted posterior. In this work, we explore a simple inference-time gradient-free guidance approach, called controlled denoising (CoDe), that circumvents the need for differentiable guidance functions and model finetuning. CoDe is a blockwise sampling method applied during intermediate denoising steps, allowing for alignment with downstream rewards. Our experiments demonstrate that, despite its simplicity, CoDe offers a favorable trad"},"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":"2502.00968","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-02-03T00:23:04Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"3906fdbbe7e423e98c153278179b9ef7b44331c2d6c659c935580fb7f1381ace","abstract_canon_sha256":"a82c9f763774b680f44da22ed0b8ae9b85390fef26fe26a0d34086561647f01c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:57:55.437664Z","signature_b64":"uGrj9CgYfnn7oy4f/CtfdSe7HOCv+asWDywoHU53Ab6BGrWkFCOMxwgMvQk9NhAj74uoDAWjkMTKFbPrgRrjAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"867bc5eab61533ca0d99b7f7c96e330d39ff6563d2b6662d82f3ca0ef1333440","last_reissued_at":"2026-07-05T10:57:55.437162Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:57:55.437162Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CoDe: Blockwise Control for Denoising Diffusion Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Ahmad Beirami, Anuj Singh, Hadi Jamali-Rad, Sayak Mukherjee","submitted_at":"2025-02-03T00:23:04Z","abstract_excerpt":"Aligning diffusion models to downstream tasks often requires finetuning new models or gradient-based guidance at inference time to enable sampling from the reward-tilted posterior. In this work, we explore a simple inference-time gradient-free guidance approach, called controlled denoising (CoDe), that circumvents the need for differentiable guidance functions and model finetuning. CoDe is a blockwise sampling method applied during intermediate denoising steps, allowing for alignment with downstream rewards. Our experiments demonstrate that, despite its simplicity, CoDe offers a favorable trad"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.00968","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/2502.00968/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":"2502.00968","created_at":"2026-07-05T10:57:55.437221+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.00968v2","created_at":"2026-07-05T10:57:55.437221+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.00968","created_at":"2026-07-05T10:57:55.437221+00:00"},{"alias_kind":"pith_short_12","alias_value":"QZ54L2VWCUZ4","created_at":"2026-07-05T10:57:55.437221+00:00"},{"alias_kind":"pith_short_16","alias_value":"QZ54L2VWCUZ4UDMZ","created_at":"2026-07-05T10:57:55.437221+00:00"},{"alias_kind":"pith_short_8","alias_value":"QZ54L2VW","created_at":"2026-07-05T10:57:55.437221+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.08231","citing_title":"Test-Time Scaling in Multimodal Foundation Models: A Comprehensive Survey of Generation and Reasoning","ref_index":71,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14379","citing_title":"Step-level Denoising-time Diffusion Alignment with Multiple Objectives","ref_index":28,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QZ54L2VWCUZ4UDMZW734S3RTBU","json":"https://pith.science/pith/QZ54L2VWCUZ4UDMZW734S3RTBU.json","graph_json":"https://pith.science/api/pith-number/QZ54L2VWCUZ4UDMZW734S3RTBU/graph.json","events_json":"https://pith.science/api/pith-number/QZ54L2VWCUZ4UDMZW734S3RTBU/events.json","paper":"https://pith.science/paper/QZ54L2VW"},"agent_actions":{"view_html":"https://pith.science/pith/QZ54L2VWCUZ4UDMZW734S3RTBU","download_json":"https://pith.science/pith/QZ54L2VWCUZ4UDMZW734S3RTBU.json","view_paper":"https://pith.science/paper/QZ54L2VW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.00968&json=true","fetch_graph":"https://pith.science/api/pith-number/QZ54L2VWCUZ4UDMZW734S3RTBU/graph.json","fetch_events":"https://pith.science/api/pith-number/QZ54L2VWCUZ4UDMZW734S3RTBU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QZ54L2VWCUZ4UDMZW734S3RTBU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QZ54L2VWCUZ4UDMZW734S3RTBU/action/storage_attestation","attest_author":"https://pith.science/pith/QZ54L2VWCUZ4UDMZW734S3RTBU/action/author_attestation","sign_citation":"https://pith.science/pith/QZ54L2VWCUZ4UDMZW734S3RTBU/action/citation_signature","submit_replication":"https://pith.science/pith/QZ54L2VWCUZ4UDMZW734S3RTBU/action/replication_record"}},"created_at":"2026-07-05T10:57:55.437221+00:00","updated_at":"2026-07-05T10:57:55.437221+00:00"}