{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:UNM2H65HH52TLX7RAEGEYQSGM3","short_pith_number":"pith:UNM2H65H","schema_version":"1.0","canonical_sha256":"a359a3fba73f7535dff1010c4c424666d97dbbda5cfd89dba5803f41e9523ff5","source":{"kind":"arxiv","id":"2502.10236","version":2},"attestation_state":"computed","paper":{"title":"Shaping Inductive Bias in Diffusion Models through Frequency-Based Noise Control","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Berton Earnshaw, Jason Hartford, Luca Scimeca, Thomas Jiralerspong, Yoshua Bengio","submitted_at":"2025-02-14T15:46:37Z","abstract_excerpt":"Diffusion Probabilistic Models (DPMs) are powerful generative models that have achieved unparalleled success in a number of generative tasks. In this work, we aim to build inductive biases into the training and sampling of diffusion models to better accommodate the target distribution of the data to model. For topologically structured data, we devise a frequency-based noising operator to purposefully manipulate, and set, these inductive biases. We first show that appropriate manipulations of the noising forward process can lead DPMs to focus on particular aspects of the distribution to learn. "},"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":"2502.10236","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-14T15:46:37Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"fc22c126df64f5a8b43aea0e7dc298f4d24d2dd76ca93ffaa33e9b4bee611bde","abstract_canon_sha256":"588a97edaee6678d94882de53e85e855586b1e7ac18d70d32c33fcaa76e73ebc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:30:03.687696Z","signature_b64":"1o0VZrhBmmtiYwv2Kjk01eRRyIJb16a5Du+FGLtt7d5KABYPHvOIRsg9NWSnLpN+PfV+kJ8zObEnEN4BYQ+NCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a359a3fba73f7535dff1010c4c424666d97dbbda5cfd89dba5803f41e9523ff5","last_reissued_at":"2026-07-05T10:30:03.686788Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:30:03.686788Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Shaping Inductive Bias in Diffusion Models through Frequency-Based Noise Control","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Berton Earnshaw, Jason Hartford, Luca Scimeca, Thomas Jiralerspong, Yoshua Bengio","submitted_at":"2025-02-14T15:46:37Z","abstract_excerpt":"Diffusion Probabilistic Models (DPMs) are powerful generative models that have achieved unparalleled success in a number of generative tasks. In this work, we aim to build inductive biases into the training and sampling of diffusion models to better accommodate the target distribution of the data to model. For topologically structured data, we devise a frequency-based noising operator to purposefully manipulate, and set, these inductive biases. We first show that appropriate manipulations of the noising forward process can lead DPMs to focus on particular aspects of the distribution to learn. "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.10236","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/2502.10236/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":"2502.10236","created_at":"2026-07-05T10:30:03.686908+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.10236v2","created_at":"2026-07-05T10:30:03.686908+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.10236","created_at":"2026-07-05T10:30:03.686908+00:00"},{"alias_kind":"pith_short_12","alias_value":"UNM2H65HH52T","created_at":"2026-07-05T10:30:03.686908+00:00"},{"alias_kind":"pith_short_16","alias_value":"UNM2H65HH52TLX7R","created_at":"2026-07-05T10:30:03.686908+00:00"},{"alias_kind":"pith_short_8","alias_value":"UNM2H65H","created_at":"2026-07-05T10:30:03.686908+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05196","citing_title":"Unified Audio Intelligence Without Regressing on Text Intelligence","ref_index":170,"is_internal_anchor":true},{"citing_arxiv_id":"2605.27919","citing_title":"Frequency-Guided Action Diffusion via Sub-Frequency Manifold Traversal","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03393","citing_title":"Flicker-DDPM: Accelerating Denoising Diffusion via 1/f Colored Noise Injection","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UNM2H65HH52TLX7RAEGEYQSGM3","json":"https://pith.science/pith/UNM2H65HH52TLX7RAEGEYQSGM3.json","graph_json":"https://pith.science/api/pith-number/UNM2H65HH52TLX7RAEGEYQSGM3/graph.json","events_json":"https://pith.science/api/pith-number/UNM2H65HH52TLX7RAEGEYQSGM3/events.json","paper":"https://pith.science/paper/UNM2H65H"},"agent_actions":{"view_html":"https://pith.science/pith/UNM2H65HH52TLX7RAEGEYQSGM3","download_json":"https://pith.science/pith/UNM2H65HH52TLX7RAEGEYQSGM3.json","view_paper":"https://pith.science/paper/UNM2H65H","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.10236&json=true","fetch_graph":"https://pith.science/api/pith-number/UNM2H65HH52TLX7RAEGEYQSGM3/graph.json","fetch_events":"https://pith.science/api/pith-number/UNM2H65HH52TLX7RAEGEYQSGM3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UNM2H65HH52TLX7RAEGEYQSGM3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UNM2H65HH52TLX7RAEGEYQSGM3/action/storage_attestation","attest_author":"https://pith.science/pith/UNM2H65HH52TLX7RAEGEYQSGM3/action/author_attestation","sign_citation":"https://pith.science/pith/UNM2H65HH52TLX7RAEGEYQSGM3/action/citation_signature","submit_replication":"https://pith.science/pith/UNM2H65HH52TLX7RAEGEYQSGM3/action/replication_record"}},"created_at":"2026-07-05T10:30:03.686908+00:00","updated_at":"2026-07-05T10:30:03.686908+00:00"}