{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:VPQYEK34MSYT5LZITPRRVICRMK","short_pith_number":"pith:VPQYEK34","schema_version":"1.0","canonical_sha256":"abe1822b7c64b13eaf289be31aa05162a6b3336bbcf4e56f0d2348d1f4aa224b","source":{"kind":"arxiv","id":"2112.06749","version":3},"attestation_state":"computed","paper":{"title":"Step-unrolled Denoising Autoencoders for Text Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Aaron van den Oord, Erich Elsen, Junyoung Chung, Mikolaj Binkowski, Nikolay Savinov","submitted_at":"2021-12-13T16:00:33Z","abstract_excerpt":"In this paper we propose a new generative model of text, Step-unrolled Denoising Autoencoder (SUNDAE), that does not rely on autoregressive models. Similarly to denoising diffusion techniques, SUNDAE is repeatedly applied on a sequence of tokens, starting from random inputs and improving them each time until convergence. We present a simple new improvement operator that converges in fewer iterations than diffusion methods, while qualitatively producing better samples on natural language datasets. SUNDAE achieves state-of-the-art results (among non-autoregressive methods) on the WMT'14 English-"},"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":"2112.06749","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-12-13T16:00:33Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"2b23f4e9c2cb9bb7e59f46791a30e1a480a59e20a1843841229dcbf6f0335037","abstract_canon_sha256":"b74ddc06f5fb905d608f032785f80c48681e2644cafe8f757f0157da6abdb102"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:15:52.433759Z","signature_b64":"IedumM97KBiazuZQvIYiaujVSg2PgwWs6PvdVxEaei2NVSNulsNH3+PhL02wpYJj9pgusBCWtVZMWkNWB09VCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"abe1822b7c64b13eaf289be31aa05162a6b3336bbcf4e56f0d2348d1f4aa224b","last_reissued_at":"2026-07-05T04:15:52.433256Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:15:52.433256Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Step-unrolled Denoising Autoencoders for Text Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Aaron van den Oord, Erich Elsen, Junyoung Chung, Mikolaj Binkowski, Nikolay Savinov","submitted_at":"2021-12-13T16:00:33Z","abstract_excerpt":"In this paper we propose a new generative model of text, Step-unrolled Denoising Autoencoder (SUNDAE), that does not rely on autoregressive models. Similarly to denoising diffusion techniques, SUNDAE is repeatedly applied on a sequence of tokens, starting from random inputs and improving them each time until convergence. We present a simple new improvement operator that converges in fewer iterations than diffusion methods, while qualitatively producing better samples on natural language datasets. SUNDAE achieves state-of-the-art results (among non-autoregressive methods) on the WMT'14 English-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2112.06749","kind":"arxiv","version":3},"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/2112.06749/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":"2112.06749","created_at":"2026-07-05T04:15:52.433323+00:00"},{"alias_kind":"arxiv_version","alias_value":"2112.06749v3","created_at":"2026-07-05T04:15:52.433323+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2112.06749","created_at":"2026-07-05T04:15:52.433323+00:00"},{"alias_kind":"pith_short_12","alias_value":"VPQYEK34MSYT","created_at":"2026-07-05T04:15:52.433323+00:00"},{"alias_kind":"pith_short_16","alias_value":"VPQYEK34MSYT5LZI","created_at":"2026-07-05T04:15:52.433323+00:00"},{"alias_kind":"pith_short_8","alias_value":"VPQYEK34","created_at":"2026-07-05T04:15:52.433323+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.08184","citing_title":"TextEconomizer: Enhancing Lossy Text Compression with Denoising Transformers and Entropy Coding","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07567","citing_title":"SurfDesign: Effective Protein Design on Molecular Surfaces","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20476","citing_title":"Goodbye Drift: Anchored Tree Sampling for Long-Horizon Video-to-Video Generation","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2211.15089","citing_title":"Continuous diffusion for categorical data","ref_index":77,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VPQYEK34MSYT5LZITPRRVICRMK","json":"https://pith.science/pith/VPQYEK34MSYT5LZITPRRVICRMK.json","graph_json":"https://pith.science/api/pith-number/VPQYEK34MSYT5LZITPRRVICRMK/graph.json","events_json":"https://pith.science/api/pith-number/VPQYEK34MSYT5LZITPRRVICRMK/events.json","paper":"https://pith.science/paper/VPQYEK34"},"agent_actions":{"view_html":"https://pith.science/pith/VPQYEK34MSYT5LZITPRRVICRMK","download_json":"https://pith.science/pith/VPQYEK34MSYT5LZITPRRVICRMK.json","view_paper":"https://pith.science/paper/VPQYEK34","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2112.06749&json=true","fetch_graph":"https://pith.science/api/pith-number/VPQYEK34MSYT5LZITPRRVICRMK/graph.json","fetch_events":"https://pith.science/api/pith-number/VPQYEK34MSYT5LZITPRRVICRMK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VPQYEK34MSYT5LZITPRRVICRMK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VPQYEK34MSYT5LZITPRRVICRMK/action/storage_attestation","attest_author":"https://pith.science/pith/VPQYEK34MSYT5LZITPRRVICRMK/action/author_attestation","sign_citation":"https://pith.science/pith/VPQYEK34MSYT5LZITPRRVICRMK/action/citation_signature","submit_replication":"https://pith.science/pith/VPQYEK34MSYT5LZITPRRVICRMK/action/replication_record"}},"created_at":"2026-07-05T04:15:52.433323+00:00","updated_at":"2026-07-05T04:15:52.433323+00:00"}