{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:47VMV7KBAISFEFFM7SRXKYTFWF","short_pith_number":"pith:47VMV7KB","schema_version":"1.0","canonical_sha256":"e7eacafd4102245214acfca3756265b1571454c6d0d60e98db4eb2c5207cad92","source":{"kind":"arxiv","id":"2008.07772","version":2},"attestation_state":"computed","paper":{"title":"Very Deep Transformers for Neural Machine Translation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jianfeng Gao, Kevin Duh, Liyuan Liu, XiaoDong Liu","submitted_at":"2020-08-18T07:14:54Z","abstract_excerpt":"We explore the application of very deep Transformer models for Neural Machine Translation (NMT). Using a simple yet effective initialization technique that stabilizes training, we show that it is feasible to build standard Transformer-based models with up to 60 encoder layers and 12 decoder layers. These deep models outperform their baseline 6-layer counterparts by as much as 2.5 BLEU, and achieve new state-of-the-art benchmark results on WMT14 English-French (43.8 BLEU and 46.4 BLEU with back-translation) and WMT14 English-German (30.1 BLEU).The code and trained models will be publicly availa"},"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":"2008.07772","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-08-18T07:14:54Z","cross_cats_sorted":[],"title_canon_sha256":"3ff9a96a3f16074fe848edfcb7e89a5c7e7239b0fe968209644a61a2220ede8b","abstract_canon_sha256":"ecdfe107692da0666f9e2c08a1528fb5eb7a5c2e9537378db3362ceac645895e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:43:05.709026Z","signature_b64":"JpS6u2T/NjU7NZZ+orhVx7kmYKPODPhkHMBCCfazZBung2Vf5Uvr4r4GTPs8WSX6RvLlcolRLD6aqCqaWkizCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e7eacafd4102245214acfca3756265b1571454c6d0d60e98db4eb2c5207cad92","last_reissued_at":"2026-07-05T01:43:05.708620Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:43:05.708620Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Very Deep Transformers for Neural Machine Translation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jianfeng Gao, Kevin Duh, Liyuan Liu, XiaoDong Liu","submitted_at":"2020-08-18T07:14:54Z","abstract_excerpt":"We explore the application of very deep Transformer models for Neural Machine Translation (NMT). Using a simple yet effective initialization technique that stabilizes training, we show that it is feasible to build standard Transformer-based models with up to 60 encoder layers and 12 decoder layers. These deep models outperform their baseline 6-layer counterparts by as much as 2.5 BLEU, and achieve new state-of-the-art benchmark results on WMT14 English-French (43.8 BLEU and 46.4 BLEU with back-translation) and WMT14 English-German (30.1 BLEU).The code and trained models will be publicly availa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2008.07772","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/2008.07772/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":"2008.07772","created_at":"2026-07-05T01:43:05.708668+00:00"},{"alias_kind":"arxiv_version","alias_value":"2008.07772v2","created_at":"2026-07-05T01:43:05.708668+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2008.07772","created_at":"2026-07-05T01:43:05.708668+00:00"},{"alias_kind":"pith_short_12","alias_value":"47VMV7KBAISF","created_at":"2026-07-05T01:43:05.708668+00:00"},{"alias_kind":"pith_short_16","alias_value":"47VMV7KBAISFEFFM","created_at":"2026-07-05T01:43:05.708668+00:00"},{"alias_kind":"pith_short_8","alias_value":"47VMV7KB","created_at":"2026-07-05T01:43:05.708668+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2306.14048","citing_title":"H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models","ref_index":86,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/47VMV7KBAISFEFFM7SRXKYTFWF","json":"https://pith.science/pith/47VMV7KBAISFEFFM7SRXKYTFWF.json","graph_json":"https://pith.science/api/pith-number/47VMV7KBAISFEFFM7SRXKYTFWF/graph.json","events_json":"https://pith.science/api/pith-number/47VMV7KBAISFEFFM7SRXKYTFWF/events.json","paper":"https://pith.science/paper/47VMV7KB"},"agent_actions":{"view_html":"https://pith.science/pith/47VMV7KBAISFEFFM7SRXKYTFWF","download_json":"https://pith.science/pith/47VMV7KBAISFEFFM7SRXKYTFWF.json","view_paper":"https://pith.science/paper/47VMV7KB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2008.07772&json=true","fetch_graph":"https://pith.science/api/pith-number/47VMV7KBAISFEFFM7SRXKYTFWF/graph.json","fetch_events":"https://pith.science/api/pith-number/47VMV7KBAISFEFFM7SRXKYTFWF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/47VMV7KBAISFEFFM7SRXKYTFWF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/47VMV7KBAISFEFFM7SRXKYTFWF/action/storage_attestation","attest_author":"https://pith.science/pith/47VMV7KBAISFEFFM7SRXKYTFWF/action/author_attestation","sign_citation":"https://pith.science/pith/47VMV7KBAISFEFFM7SRXKYTFWF/action/citation_signature","submit_replication":"https://pith.science/pith/47VMV7KBAISFEFFM7SRXKYTFWF/action/replication_record"}},"created_at":"2026-07-05T01:43:05.708668+00:00","updated_at":"2026-07-05T01:43:05.708668+00:00"}