{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:GHK4S6CNGXVGKRIIPUNNZSDAKT","short_pith_number":"pith:GHK4S6CN","schema_version":"1.0","canonical_sha256":"31d5c9784d35ea6545087d1adcc86054cf3f0d450346b83ad171942b5dc0d168","source":{"kind":"arxiv","id":"2506.19713","version":1},"attestation_state":"computed","paper":{"title":"Guidance in the Frequency Domain Enables High-Fidelity Sampling at Low CFG Scales","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Farnood Salehi, Romann M. Weber, Seyedmorteza Sadat, Tobias Vontobel","submitted_at":"2025-06-24T15:19:42Z","abstract_excerpt":"Classifier-free guidance (CFG) has become an essential component of modern conditional diffusion models. Although highly effective in practice, the underlying mechanisms by which CFG enhances quality, detail, and prompt alignment are not fully understood. We present a novel perspective on CFG by analyzing its effects in the frequency domain, showing that low and high frequencies have distinct impacts on generation quality. Specifically, low-frequency guidance governs global structure and condition alignment, while high-frequency guidance mainly enhances visual fidelity. However, applying a uni"},"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":"2506.19713","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-06-24T15:19:42Z","cross_cats_sorted":[],"title_canon_sha256":"d9fe42e2c8337d7e22ff5c33051bc3fb923fd12f66ae45b2dd57952e03b91230","abstract_canon_sha256":"ba4ba97bf1a300eba04c997e221144ab08609a953eda83b06f892b09dbf4e0a7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:26:37.586142Z","signature_b64":"YFxH7BYbcJ9RZwOTD23V7aNY/cryBW4p3Bl+7F/klfUmasLW1ZXWv2qcVq6HMf+RZBwIhV0iTFo9evuQX+7XCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"31d5c9784d35ea6545087d1adcc86054cf3f0d450346b83ad171942b5dc0d168","last_reissued_at":"2026-07-05T11:26:37.585650Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:26:37.585650Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Guidance in the Frequency Domain Enables High-Fidelity Sampling at Low CFG Scales","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Farnood Salehi, Romann M. Weber, Seyedmorteza Sadat, Tobias Vontobel","submitted_at":"2025-06-24T15:19:42Z","abstract_excerpt":"Classifier-free guidance (CFG) has become an essential component of modern conditional diffusion models. Although highly effective in practice, the underlying mechanisms by which CFG enhances quality, detail, and prompt alignment are not fully understood. We present a novel perspective on CFG by analyzing its effects in the frequency domain, showing that low and high frequencies have distinct impacts on generation quality. Specifically, low-frequency guidance governs global structure and condition alignment, while high-frequency guidance mainly enhances visual fidelity. However, applying a uni"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.19713","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/2506.19713/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":"2506.19713","created_at":"2026-07-05T11:26:37.585709+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.19713v1","created_at":"2026-07-05T11:26:37.585709+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.19713","created_at":"2026-07-05T11:26:37.585709+00:00"},{"alias_kind":"pith_short_12","alias_value":"GHK4S6CNGXVG","created_at":"2026-07-05T11:26:37.585709+00:00"},{"alias_kind":"pith_short_16","alias_value":"GHK4S6CNGXVGKRII","created_at":"2026-07-05T11:26:37.585709+00:00"},{"alias_kind":"pith_short_8","alias_value":"GHK4S6CN","created_at":"2026-07-05T11:26:37.585709+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18765","citing_title":"SpectralDiT: Timestep-Conditioned Spectral Residual Correction for Flow-Matching DiTs","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2603.08155","citing_title":"C$^2$FG: Control Classifier-Free Guidance via Score Discrepancy Analysis","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2603.08155","citing_title":"C$^2$FG: Control Classifier-Free Guidance via Score Discrepancy Analysis","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13910","citing_title":"Covariance-aware sampling for Diffusion Models","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GHK4S6CNGXVGKRIIPUNNZSDAKT","json":"https://pith.science/pith/GHK4S6CNGXVGKRIIPUNNZSDAKT.json","graph_json":"https://pith.science/api/pith-number/GHK4S6CNGXVGKRIIPUNNZSDAKT/graph.json","events_json":"https://pith.science/api/pith-number/GHK4S6CNGXVGKRIIPUNNZSDAKT/events.json","paper":"https://pith.science/paper/GHK4S6CN"},"agent_actions":{"view_html":"https://pith.science/pith/GHK4S6CNGXVGKRIIPUNNZSDAKT","download_json":"https://pith.science/pith/GHK4S6CNGXVGKRIIPUNNZSDAKT.json","view_paper":"https://pith.science/paper/GHK4S6CN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.19713&json=true","fetch_graph":"https://pith.science/api/pith-number/GHK4S6CNGXVGKRIIPUNNZSDAKT/graph.json","fetch_events":"https://pith.science/api/pith-number/GHK4S6CNGXVGKRIIPUNNZSDAKT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GHK4S6CNGXVGKRIIPUNNZSDAKT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GHK4S6CNGXVGKRIIPUNNZSDAKT/action/storage_attestation","attest_author":"https://pith.science/pith/GHK4S6CNGXVGKRIIPUNNZSDAKT/action/author_attestation","sign_citation":"https://pith.science/pith/GHK4S6CNGXVGKRIIPUNNZSDAKT/action/citation_signature","submit_replication":"https://pith.science/pith/GHK4S6CNGXVGKRIIPUNNZSDAKT/action/replication_record"}},"created_at":"2026-07-05T11:26:37.585709+00:00","updated_at":"2026-07-05T11:26:37.585709+00:00"}