{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:HPRJRRSWLHFTSR4EGOKNH23ME3","short_pith_number":"pith:HPRJRRSW","schema_version":"1.0","canonical_sha256":"3be298c65659cb3947843394d3eb6c26c871b5815fb64f9a65dd91a3acb61af9","source":{"kind":"arxiv","id":"2206.00330","version":2},"attestation_state":"computed","paper":{"title":"B2T Connection: Serving Stability and Performance in Deep Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Jun Suzuki, Sho Takase, Shun Kiyono, Sosuke Kobayashi","submitted_at":"2022-06-01T08:43:20Z","abstract_excerpt":"From the perspective of the layer normalization (LN) positions, the architectures of Transformers can be categorized into two types: Post-LN and Pre-LN. Recent Transformers tend to be Pre-LN because, in Post-LN with deep Transformers (e.g., those with ten or more layers), the training is often unstable, resulting in useless models. However, Post-LN has consistently achieved better performance than Pre-LN in relatively shallow Transformers (e.g., those with six or fewer layers). This study first investigates the reason for these discrepant observations empirically and theoretically and made the"},"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":"2206.00330","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-06-01T08:43:20Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"a8ef8072981ce9ce771a6d0ab441684c34bc9208aea357873f8a3111122499f2","abstract_canon_sha256":"30e3dca0a1fde90cccf64c7670b93aaa0df00fd92350cf5d507ffa4ec3f718f4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:14:04.188338Z","signature_b64":"YstMkZth3n5NDIVMRjeHlfkYufqhAT0q/xD8RJhPo5EqjHni5GcUXjCJcm71MHPMdd2m9/92v6JoJCVWiQIRBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3be298c65659cb3947843394d3eb6c26c871b5815fb64f9a65dd91a3acb61af9","last_reissued_at":"2026-07-05T06:14:04.187893Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:14:04.187893Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"B2T Connection: Serving Stability and Performance in Deep Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Jun Suzuki, Sho Takase, Shun Kiyono, Sosuke Kobayashi","submitted_at":"2022-06-01T08:43:20Z","abstract_excerpt":"From the perspective of the layer normalization (LN) positions, the architectures of Transformers can be categorized into two types: Post-LN and Pre-LN. Recent Transformers tend to be Pre-LN because, in Post-LN with deep Transformers (e.g., those with ten or more layers), the training is often unstable, resulting in useless models. However, Post-LN has consistently achieved better performance than Pre-LN in relatively shallow Transformers (e.g., those with six or fewer layers). This study first investigates the reason for these discrepant observations empirically and theoretically and made the"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2206.00330","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/2206.00330/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":"2206.00330","created_at":"2026-07-05T06:14:04.187948+00:00"},{"alias_kind":"arxiv_version","alias_value":"2206.00330v2","created_at":"2026-07-05T06:14:04.187948+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2206.00330","created_at":"2026-07-05T06:14:04.187948+00:00"},{"alias_kind":"pith_short_12","alias_value":"HPRJRRSWLHFT","created_at":"2026-07-05T06:14:04.187948+00:00"},{"alias_kind":"pith_short_16","alias_value":"HPRJRRSWLHFTSR4E","created_at":"2026-07-05T06:14:04.187948+00:00"},{"alias_kind":"pith_short_8","alias_value":"HPRJRRSW","created_at":"2026-07-05T06:14:04.187948+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26538","citing_title":"CascadeFormer: Depth-Tapered Transformers Motivated by Gradient Fan-in Asymmetry","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18389","citing_title":"Understanding the Prompt Sensitivity","ref_index":29,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HPRJRRSWLHFTSR4EGOKNH23ME3","json":"https://pith.science/pith/HPRJRRSWLHFTSR4EGOKNH23ME3.json","graph_json":"https://pith.science/api/pith-number/HPRJRRSWLHFTSR4EGOKNH23ME3/graph.json","events_json":"https://pith.science/api/pith-number/HPRJRRSWLHFTSR4EGOKNH23ME3/events.json","paper":"https://pith.science/paper/HPRJRRSW"},"agent_actions":{"view_html":"https://pith.science/pith/HPRJRRSWLHFTSR4EGOKNH23ME3","download_json":"https://pith.science/pith/HPRJRRSWLHFTSR4EGOKNH23ME3.json","view_paper":"https://pith.science/paper/HPRJRRSW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2206.00330&json=true","fetch_graph":"https://pith.science/api/pith-number/HPRJRRSWLHFTSR4EGOKNH23ME3/graph.json","fetch_events":"https://pith.science/api/pith-number/HPRJRRSWLHFTSR4EGOKNH23ME3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HPRJRRSWLHFTSR4EGOKNH23ME3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HPRJRRSWLHFTSR4EGOKNH23ME3/action/storage_attestation","attest_author":"https://pith.science/pith/HPRJRRSWLHFTSR4EGOKNH23ME3/action/author_attestation","sign_citation":"https://pith.science/pith/HPRJRRSWLHFTSR4EGOKNH23ME3/action/citation_signature","submit_replication":"https://pith.science/pith/HPRJRRSWLHFTSR4EGOKNH23ME3/action/replication_record"}},"created_at":"2026-07-05T06:14:04.187948+00:00","updated_at":"2026-07-05T06:14:04.187948+00:00"}