{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:NLVIECDOXB5FZVOD2FTYZJ6XYS","short_pith_number":"pith:NLVIECDO","schema_version":"1.0","canonical_sha256":"6aea82086eb87a5cd5c3d1678ca7d7c49326c83e674f0b749b7a0713df93aec5","source":{"kind":"arxiv","id":"2605.01817","version":2},"attestation_state":"computed","paper":{"title":"Skipping the Zeros in Diffusion Models for Sparse Data Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Diffusion models can generate sparse data by modeling only non-zero values while handling zero locations separately.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andriy Balinskyy, Carl Herrmann, Gabriel Vicente Rodrigues, Jean Radig, Marius Kloft, Mayank Nagda, Phil Sidney Ostheimer, Sophie Fellenz, Stephan Mandt","submitted_at":"2026-05-03T10:51:25Z","abstract_excerpt":"Diffusion models (DMs) excel on dense continuous data, but are not designed for sparse continuous data. They do not model exact zeros that represent the deliberate absence of a signal. As a result, they erase sparsity patterns and perform unnecessary computation on mostly zero entries. With Sparsity-Exploiting Diffusion (SED), we model only non-zero values, preserving sparsity. SED delivers computational savings while maintaining or improving generation quality by skipping zeros during training and inference. Across physics and biology benchmarks, SED matches or surpasses conventional DMs and "},"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":"2605.01817","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-03T10:51:25Z","cross_cats_sorted":[],"title_canon_sha256":"1dd872630ceb03138578a41ab6e53156a7d63d0b85f58bce000883750c5ff7a2","abstract_canon_sha256":"a8d46a7d4c95a6441c9ed13d0970d4453b79cc421f31eff3f953af06934ee603"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T01:05:55.830047Z","signature_b64":"Una1V/tqGYk62v1UwkEH8BTqAgT+d8BmuLngL8hi5fxuTMpXUO/E4QegM+Sw1eLhZ9A2rJq2KoqJuz3m1UWBCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6aea82086eb87a5cd5c3d1678ca7d7c49326c83e674f0b749b7a0713df93aec5","last_reissued_at":"2026-05-27T01:05:55.829130Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T01:05:55.829130Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Skipping the Zeros in Diffusion Models for Sparse Data Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Diffusion models can generate sparse data by modeling only non-zero values while handling zero locations separately.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andriy Balinskyy, Carl Herrmann, Gabriel Vicente Rodrigues, Jean Radig, Marius Kloft, Mayank Nagda, Phil Sidney Ostheimer, Sophie Fellenz, Stephan Mandt","submitted_at":"2026-05-03T10:51:25Z","abstract_excerpt":"Diffusion models (DMs) excel on dense continuous data, but are not designed for sparse continuous data. They do not model exact zeros that represent the deliberate absence of a signal. As a result, they erase sparsity patterns and perform unnecessary computation on mostly zero entries. With Sparsity-Exploiting Diffusion (SED), we model only non-zero values, preserving sparsity. SED delivers computational savings while maintaining or improving generation quality by skipping zeros during training and inference. Across physics and biology benchmarks, SED matches or surpasses conventional DMs and "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"With Sparsity-Exploiting Diffusion (SED), we model only non-zero values, preserving sparsity. SED delivers computational savings while maintaining or improving generation quality by skipping zeros during training and inference.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the sparsity pattern (locations of zeros) is independent of the non-zero values and can be handled separately without losing critical distributional information about the data.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SED modifies diffusion models to generate only non-zero values in sparse data, preserving sparsity patterns, cutting computation, and matching or beating standard DM performance on benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Diffusion models can generate sparse data by modeling only non-zero values while handling zero locations separately.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"62dde6ef9f86f97a00c9c099ee095ed21f0f19bf0731026713d3fbe91be512ca"},"source":{"id":"2605.01817","kind":"arxiv","version":2},"verdict":{"id":"a538bf93-d483-4aa9-ae77-a3f7e7ddc1a8","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T14:53:16.044252Z","strongest_claim":"With Sparsity-Exploiting Diffusion (SED), we model only non-zero values, preserving sparsity. SED delivers computational savings while maintaining or improving generation quality by skipping zeros during training and inference.","one_line_summary":"SED modifies diffusion models to generate only non-zero values in sparse data, preserving sparsity patterns, cutting computation, and matching or beating standard DM performance on benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the sparsity pattern (locations of zeros) is independent of the non-zero values and can be handled separately without losing critical distributional information about the data.","pith_extraction_headline":"Diffusion models can generate sparse data by modeling only non-zero values while handling zero locations separately."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.01817/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T17:35:39.933380Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-20T05:01:22.604083Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T16:57:51.717741Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"13a0102c790b16546b96f0b1fc3fc02505166a419dabe84745bd55be3eaec540"},"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":"2605.01817","created_at":"2026-05-27T01:05:55.829220+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.01817v2","created_at":"2026-05-27T01:05:55.829220+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.01817","created_at":"2026-05-27T01:05:55.829220+00:00"},{"alias_kind":"pith_short_12","alias_value":"NLVIECDOXB5F","created_at":"2026-05-27T01:05:55.829220+00:00"},{"alias_kind":"pith_short_16","alias_value":"NLVIECDOXB5FZVOD","created_at":"2026-05-27T01:05:55.829220+00:00"},{"alias_kind":"pith_short_8","alias_value":"NLVIECDO","created_at":"2026-05-27T01:05:55.829220+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NLVIECDOXB5FZVOD2FTYZJ6XYS","json":"https://pith.science/pith/NLVIECDOXB5FZVOD2FTYZJ6XYS.json","graph_json":"https://pith.science/api/pith-number/NLVIECDOXB5FZVOD2FTYZJ6XYS/graph.json","events_json":"https://pith.science/api/pith-number/NLVIECDOXB5FZVOD2FTYZJ6XYS/events.json","paper":"https://pith.science/paper/NLVIECDO"},"agent_actions":{"view_html":"https://pith.science/pith/NLVIECDOXB5FZVOD2FTYZJ6XYS","download_json":"https://pith.science/pith/NLVIECDOXB5FZVOD2FTYZJ6XYS.json","view_paper":"https://pith.science/paper/NLVIECDO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.01817&json=true","fetch_graph":"https://pith.science/api/pith-number/NLVIECDOXB5FZVOD2FTYZJ6XYS/graph.json","fetch_events":"https://pith.science/api/pith-number/NLVIECDOXB5FZVOD2FTYZJ6XYS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NLVIECDOXB5FZVOD2FTYZJ6XYS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NLVIECDOXB5FZVOD2FTYZJ6XYS/action/storage_attestation","attest_author":"https://pith.science/pith/NLVIECDOXB5FZVOD2FTYZJ6XYS/action/author_attestation","sign_citation":"https://pith.science/pith/NLVIECDOXB5FZVOD2FTYZJ6XYS/action/citation_signature","submit_replication":"https://pith.science/pith/NLVIECDOXB5FZVOD2FTYZJ6XYS/action/replication_record"}},"created_at":"2026-05-27T01:05:55.829220+00:00","updated_at":"2026-05-27T01:05:55.829220+00:00"}