{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:HKACEKVMBXLKR4JNYS3XI26WSM","short_pith_number":"pith:HKACEKVM","schema_version":"1.0","canonical_sha256":"3a80222aac0dd6a8f12dc4b7746bd6932f4fd01fa9394e52867d55fd0c132bfd","source":{"kind":"arxiv","id":"2203.00555","version":1},"attestation_state":"computed","paper":{"title":"DeepNet: Scaling Transformers to 1,000 Layers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Dongdong Zhang, Furu Wei, Hongyu Wang, Li Dong, Shaohan Huang, Shuming Ma","submitted_at":"2022-03-01T15:36:38Z","abstract_excerpt":"In this paper, we propose a simple yet effective method to stabilize extremely deep Transformers. Specifically, we introduce a new normalization function (DeepNorm) to modify the residual connection in Transformer, accompanying with theoretically derived initialization. In-depth theoretical analysis shows that model updates can be bounded in a stable way. The proposed method combines the best of two worlds, i.e., good performance of Post-LN and stable training of Pre-LN, making DeepNorm a preferred alternative. We successfully scale Transformers up to 1,000 layers (i.e., 2,500 attention and fe"},"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":"2203.00555","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-03-01T15:36:38Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"44c780711ca0898d8eb7326fa7123d14e5d72f78aedb03db30c7f742aa2f8107","abstract_canon_sha256":"80e3417bf8fab4e2cafa37804df467f0fc5c79efd29e43532abc2c298a2c87db"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:01:18.520908Z","signature_b64":"+98BfqP4Ad1a0mVSgGXETmHBjhPkGd6QKwouyE9X0ksMjvKJeSWkYu/hnTnlq1JsVFyjKLU3jpOTAR7eYI56CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3a80222aac0dd6a8f12dc4b7746bd6932f4fd01fa9394e52867d55fd0c132bfd","last_reissued_at":"2026-07-05T04:01:18.520429Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:01:18.520429Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DeepNet: Scaling Transformers to 1,000 Layers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Dongdong Zhang, Furu Wei, Hongyu Wang, Li Dong, Shaohan Huang, Shuming Ma","submitted_at":"2022-03-01T15:36:38Z","abstract_excerpt":"In this paper, we propose a simple yet effective method to stabilize extremely deep Transformers. Specifically, we introduce a new normalization function (DeepNorm) to modify the residual connection in Transformer, accompanying with theoretically derived initialization. In-depth theoretical analysis shows that model updates can be bounded in a stable way. The proposed method combines the best of two worlds, i.e., good performance of Post-LN and stable training of Pre-LN, making DeepNorm a preferred alternative. We successfully scale Transformers up to 1,000 layers (i.e., 2,500 attention and fe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2203.00555","kind":"arxiv","version":1},"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/2203.00555/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":"2203.00555","created_at":"2026-07-05T04:01:18.520488+00:00"},{"alias_kind":"arxiv_version","alias_value":"2203.00555v1","created_at":"2026-07-05T04:01:18.520488+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2203.00555","created_at":"2026-07-05T04:01:18.520488+00:00"},{"alias_kind":"pith_short_12","alias_value":"HKACEKVMBXLK","created_at":"2026-07-05T04:01:18.520488+00:00"},{"alias_kind":"pith_short_16","alias_value":"HKACEKVMBXLKR4JN","created_at":"2026-07-05T04:01:18.520488+00:00"},{"alias_kind":"pith_short_8","alias_value":"HKACEKVM","created_at":"2026-07-05T04:01:18.520488+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":15,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24898","citing_title":"Dense Supervision Is Not Enough: The Readout Blind Spot in Looped Language Models","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06564","citing_title":"HAARES Half-Split Residual Basis Routing for Deep Transformers","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31859","citing_title":"Review Residuals: Update-Conditioned Residual Gating for Transformers","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.26538","citing_title":"CascadeFormer: Depth-Tapered Transformers Motivated by Gradient Fan-in Asymmetry","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2602.10408","citing_title":"Gated Normalization Removal and Scale Anchoring in Pre-Norm Transformers","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2603.15031","citing_title":"Attention Residuals","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18855","citing_title":"Delta Attention Residuals","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18878","citing_title":"Prognostic Value of Lung Ultrasound Biomarkers for Readmission Risk in Congestive Heart Failure: A Pilot Data-Driven Analysis","ref_index":274,"is_internal_anchor":false},{"citing_arxiv_id":"2307.06435","citing_title":"A Comprehensive Overview of Large Language Models","ref_index":79,"is_internal_anchor":false},{"citing_arxiv_id":"2306.14048","citing_title":"H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models","ref_index":90,"is_internal_anchor":false},{"citing_arxiv_id":"2302.14045","citing_title":"Language Is Not All You Need: Aligning Perception with Language Models","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2205.14135","citing_title":"FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness","ref_index":83,"is_internal_anchor":false},{"citing_arxiv_id":"2505.06708","citing_title":"Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2307.08621","citing_title":"Retentive Network: A Successor to Transformer for Large Language Models","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2303.18223","citing_title":"A Survey of Large Language Models","ref_index":282,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HKACEKVMBXLKR4JNYS3XI26WSM","json":"https://pith.science/pith/HKACEKVMBXLKR4JNYS3XI26WSM.json","graph_json":"https://pith.science/api/pith-number/HKACEKVMBXLKR4JNYS3XI26WSM/graph.json","events_json":"https://pith.science/api/pith-number/HKACEKVMBXLKR4JNYS3XI26WSM/events.json","paper":"https://pith.science/paper/HKACEKVM"},"agent_actions":{"view_html":"https://pith.science/pith/HKACEKVMBXLKR4JNYS3XI26WSM","download_json":"https://pith.science/pith/HKACEKVMBXLKR4JNYS3XI26WSM.json","view_paper":"https://pith.science/paper/HKACEKVM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2203.00555&json=true","fetch_graph":"https://pith.science/api/pith-number/HKACEKVMBXLKR4JNYS3XI26WSM/graph.json","fetch_events":"https://pith.science/api/pith-number/HKACEKVMBXLKR4JNYS3XI26WSM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HKACEKVMBXLKR4JNYS3XI26WSM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HKACEKVMBXLKR4JNYS3XI26WSM/action/storage_attestation","attest_author":"https://pith.science/pith/HKACEKVMBXLKR4JNYS3XI26WSM/action/author_attestation","sign_citation":"https://pith.science/pith/HKACEKVMBXLKR4JNYS3XI26WSM/action/citation_signature","submit_replication":"https://pith.science/pith/HKACEKVMBXLKR4JNYS3XI26WSM/action/replication_record"}},"created_at":"2026-07-05T04:01:18.520488+00:00","updated_at":"2026-07-05T04:01:18.520488+00:00"}