{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:ACOHDHWEAOJLFFBYXHPNNS3NDE","short_pith_number":"pith:ACOHDHWE","schema_version":"1.0","canonical_sha256":"009c719ec40392b29438b9ded6cb6d193b20f988aa96f7412afcf856f6a58a57","source":{"kind":"arxiv","id":"1808.04444","version":2},"attestation_state":"computed","paper":{"title":"Character-Level Language Modeling with Deeper Self-Attention","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","stat.ML"],"primary_cat":"cs.CL","authors_text":"Dokook Choe, Llion Jones, Mandy Guo, Noah Constant, Rami Al-Rfou","submitted_at":"2018-08-09T18:44:38Z","abstract_excerpt":"LSTMs and other RNN variants have shown strong performance on character-level language modeling. These models are typically trained using truncated backpropagation through time, and it is common to assume that their success stems from their ability to remember long-term contexts. In this paper, we show that a deep (64-layer) transformer model with fixed context outperforms RNN variants by a large margin, achieving state of the art on two popular benchmarks: 1.13 bits per character on text8 and 1.06 on enwik8. To get good results at this depth, we show that it is important to add auxiliary loss"},"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":"1808.04444","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2018-08-09T18:44:38Z","cross_cats_sorted":["cs.AI","cs.LG","stat.ML"],"title_canon_sha256":"223c492c6cb0e0e1af30c360e52c22884c14521a00716ba00c60a0ca9271f543","abstract_canon_sha256":"244e384bc2171ebd2e9dae3bae885e15ac0f77e2edaaeb743e4fac84a9d66f90"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:58:47.692660Z","signature_b64":"eXy+wFFtM07v/jyDCdkJ+8lFRTzOIWEf4MTU0PfM8VhLwW5CxAVhZXV5Oq4jARYJlDKkGhdj1+5NFS5DuuP5AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"009c719ec40392b29438b9ded6cb6d193b20f988aa96f7412afcf856f6a58a57","last_reissued_at":"2026-05-17T23:58:47.692179Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:58:47.692179Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Character-Level Language Modeling with Deeper Self-Attention","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","stat.ML"],"primary_cat":"cs.CL","authors_text":"Dokook Choe, Llion Jones, Mandy Guo, Noah Constant, Rami Al-Rfou","submitted_at":"2018-08-09T18:44:38Z","abstract_excerpt":"LSTMs and other RNN variants have shown strong performance on character-level language modeling. These models are typically trained using truncated backpropagation through time, and it is common to assume that their success stems from their ability to remember long-term contexts. In this paper, we show that a deep (64-layer) transformer model with fixed context outperforms RNN variants by a large margin, achieving state of the art on two popular benchmarks: 1.13 bits per character on text8 and 1.06 on enwik8. To get good results at this depth, we show that it is important to add auxiliary loss"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1808.04444","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":""},"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":"1808.04444","created_at":"2026-05-17T23:58:47.692257+00:00"},{"alias_kind":"arxiv_version","alias_value":"1808.04444v2","created_at":"2026-05-17T23:58:47.692257+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1808.04444","created_at":"2026-05-17T23:58:47.692257+00:00"},{"alias_kind":"pith_short_12","alias_value":"ACOHDHWEAOJL","created_at":"2026-05-18T12:32:13.499390+00:00"},{"alias_kind":"pith_short_16","alias_value":"ACOHDHWEAOJLFFBY","created_at":"2026-05-18T12:32:13.499390+00:00"},{"alias_kind":"pith_short_8","alias_value":"ACOHDHWE","created_at":"2026-05-18T12:32:13.499390+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"1907.05572","citing_title":"R-Transformer: Recurrent Neural Network Enhanced Transformer","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"1906.08237","citing_title":"XLNet: Generalized Autoregressive Pretraining for Language Understanding","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2001.04451","citing_title":"Reformer: The Efficient Transformer","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"1904.10509","citing_title":"Generating Long Sequences with Sparse Transformers","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ACOHDHWEAOJLFFBYXHPNNS3NDE","json":"https://pith.science/pith/ACOHDHWEAOJLFFBYXHPNNS3NDE.json","graph_json":"https://pith.science/api/pith-number/ACOHDHWEAOJLFFBYXHPNNS3NDE/graph.json","events_json":"https://pith.science/api/pith-number/ACOHDHWEAOJLFFBYXHPNNS3NDE/events.json","paper":"https://pith.science/paper/ACOHDHWE"},"agent_actions":{"view_html":"https://pith.science/pith/ACOHDHWEAOJLFFBYXHPNNS3NDE","download_json":"https://pith.science/pith/ACOHDHWEAOJLFFBYXHPNNS3NDE.json","view_paper":"https://pith.science/paper/ACOHDHWE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1808.04444&json=true","fetch_graph":"https://pith.science/api/pith-number/ACOHDHWEAOJLFFBYXHPNNS3NDE/graph.json","fetch_events":"https://pith.science/api/pith-number/ACOHDHWEAOJLFFBYXHPNNS3NDE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ACOHDHWEAOJLFFBYXHPNNS3NDE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ACOHDHWEAOJLFFBYXHPNNS3NDE/action/storage_attestation","attest_author":"https://pith.science/pith/ACOHDHWEAOJLFFBYXHPNNS3NDE/action/author_attestation","sign_citation":"https://pith.science/pith/ACOHDHWEAOJLFFBYXHPNNS3NDE/action/citation_signature","submit_replication":"https://pith.science/pith/ACOHDHWEAOJLFFBYXHPNNS3NDE/action/replication_record"}},"created_at":"2026-05-17T23:58:47.692257+00:00","updated_at":"2026-05-17T23:58:47.692257+00:00"}