{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:MUCVAEPT5VHZST4TQQKSB4PSXH","short_pith_number":"pith:MUCVAEPT","schema_version":"1.0","canonical_sha256":"65055011f3ed4f994f93841520f1f2b9ed1f9a983933aa9bc326fbfb62531bfc","source":{"kind":"arxiv","id":"1803.03382","version":6},"attestation_state":"computed","paper":{"title":"Fast Decoding in Sequence Models using Discrete Latent Variables","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ashish Vaswani, Aurko Roy, Jakob Uszkoreit, {\\L}ukasz Kaiser, Niki Parmar, Noam Shazeer, Samy Bengio","submitted_at":"2018-03-09T04:39:35Z","abstract_excerpt":"Autoregressive sequence models based on deep neural networks, such as RNNs, Wavenet and the Transformer attain state-of-the-art results on many tasks. However, they are difficult to parallelize and are thus slow at processing long sequences. RNNs lack parallelism both during training and decoding, while architectures like WaveNet and Transformer are much more parallelizable during training, yet still operate sequentially during decoding.\n  Inspired by [arxiv:1711.00937], we present a method to extend sequence models using discrete latent variables that makes decoding much more parallelizable. "},"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":"1803.03382","kind":"arxiv","version":6},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2018-03-09T04:39:35Z","cross_cats_sorted":[],"title_canon_sha256":"0845ef0325b0410522a5aff1ba3d309af76c079df70936d28197435aee3f78b7","abstract_canon_sha256":"727571f5ba9c974a6eec8c3bf47b9396392482f850ca6b1d08175bf3b8cfe22e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:13:54.941245Z","signature_b64":"qPGpf4XxEXd1386xi0BXePM88dRpL6x06oROsbeaP1Pnd2O45xvBOgbzJ3OtOKgueh9MZ0BVe30dW16MDZeKDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"65055011f3ed4f994f93841520f1f2b9ed1f9a983933aa9bc326fbfb62531bfc","last_reissued_at":"2026-05-18T00:13:54.940604Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:13:54.940604Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fast Decoding in Sequence Models using Discrete Latent Variables","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ashish Vaswani, Aurko Roy, Jakob Uszkoreit, {\\L}ukasz Kaiser, Niki Parmar, Noam Shazeer, Samy Bengio","submitted_at":"2018-03-09T04:39:35Z","abstract_excerpt":"Autoregressive sequence models based on deep neural networks, such as RNNs, Wavenet and the Transformer attain state-of-the-art results on many tasks. However, they are difficult to parallelize and are thus slow at processing long sequences. RNNs lack parallelism both during training and decoding, while architectures like WaveNet and Transformer are much more parallelizable during training, yet still operate sequentially during decoding.\n  Inspired by [arxiv:1711.00937], we present a method to extend sequence models using discrete latent variables that makes decoding much more parallelizable. "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1803.03382","kind":"arxiv","version":6},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"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":"1803.03382","created_at":"2026-05-18T00:13:54.940705+00:00"},{"alias_kind":"arxiv_version","alias_value":"1803.03382v6","created_at":"2026-05-18T00:13:54.940705+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1803.03382","created_at":"2026-05-18T00:13:54.940705+00:00"},{"alias_kind":"pith_short_12","alias_value":"MUCVAEPT5VHZ","created_at":"2026-05-18T12:32:40.477152+00:00"},{"alias_kind":"pith_short_16","alias_value":"MUCVAEPT5VHZST4T","created_at":"2026-05-18T12:32:40.477152+00:00"},{"alias_kind":"pith_short_8","alias_value":"MUCVAEPT","created_at":"2026-05-18T12:32:40.477152+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"1906.09444","citing_title":"Retrieving Sequential Information for Non-Autoregressive Neural Machine Translation","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"1906.09601","citing_title":"Sequence Generation: From Both Sides to the Middle","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"1909.05858","citing_title":"CTRL: A Conditional Transformer Language Model for Controllable Generation","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2512.03643","citing_title":"Optical Context Compression Is Just (Bad) Autoencoding","ref_index":6,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MUCVAEPT5VHZST4TQQKSB4PSXH","json":"https://pith.science/pith/MUCVAEPT5VHZST4TQQKSB4PSXH.json","graph_json":"https://pith.science/api/pith-number/MUCVAEPT5VHZST4TQQKSB4PSXH/graph.json","events_json":"https://pith.science/api/pith-number/MUCVAEPT5VHZST4TQQKSB4PSXH/events.json","paper":"https://pith.science/paper/MUCVAEPT"},"agent_actions":{"view_html":"https://pith.science/pith/MUCVAEPT5VHZST4TQQKSB4PSXH","download_json":"https://pith.science/pith/MUCVAEPT5VHZST4TQQKSB4PSXH.json","view_paper":"https://pith.science/paper/MUCVAEPT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1803.03382&json=true","fetch_graph":"https://pith.science/api/pith-number/MUCVAEPT5VHZST4TQQKSB4PSXH/graph.json","fetch_events":"https://pith.science/api/pith-number/MUCVAEPT5VHZST4TQQKSB4PSXH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MUCVAEPT5VHZST4TQQKSB4PSXH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MUCVAEPT5VHZST4TQQKSB4PSXH/action/storage_attestation","attest_author":"https://pith.science/pith/MUCVAEPT5VHZST4TQQKSB4PSXH/action/author_attestation","sign_citation":"https://pith.science/pith/MUCVAEPT5VHZST4TQQKSB4PSXH/action/citation_signature","submit_replication":"https://pith.science/pith/MUCVAEPT5VHZST4TQQKSB4PSXH/action/replication_record"}},"created_at":"2026-05-18T00:13:54.940705+00:00","updated_at":"2026-05-18T00:13:54.940705+00:00"}