{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:PHQ7GOGAAQRW2DV6AVSM2XPZ5R","short_pith_number":"pith:PHQ7GOGA","schema_version":"1.0","canonical_sha256":"79e1f338c004236d0ebe0564cd5df9ec7b352524b96c57019714bd887d4b302a","source":{"kind":"arxiv","id":"2502.02732","version":3},"attestation_state":"computed","paper":{"title":"Peri-LN: Revisiting Normalization Layer in the Transformer Architecture","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Beomjun Kim, Byeongchan Lee, Cheonbok Park, Dongyoon Han, Jeonghoon Kim, Jinwoo Shin, Kang Min Yoo, Seongjin Shin, Taehwan Yoo, Yeontaek Oh","submitted_at":"2025-02-04T21:29:47Z","abstract_excerpt":"Selecting a layer normalization (LN) strategy that stabilizes training and speeds convergence in Transformers remains difficult, even for today's large language models (LLM). We present a comprehensive analytical foundation for understanding how different LN strategies influence training dynamics in large-scale Transformers. Until recently, Pre-LN and Post-LN have long dominated practices despite their limitations in large-scale training. However, several open-source models have recently begun silently adopting a third strategy without much explanation. This strategy places normalization layer"},"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":"2502.02732","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-04T21:29:47Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"0923a0631cc89f621895637be75b2ab0b1d5a60920801a3e4091d0b295b7dec5","abstract_canon_sha256":"0deea64e226037978aa5448913f18b5c27cb0c7787052741c990bf4d05b06af1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:16:54.732490Z","signature_b64":"PvI36A/mRP4X/kGF8klSi/Xx8GoKdk89fp4aR47MobuOW0gVE0IUlujoDCZj7mY1rTzLKm8MKQrOYmKh9l9iAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"79e1f338c004236d0ebe0564cd5df9ec7b352524b96c57019714bd887d4b302a","last_reissued_at":"2026-07-05T11:16:54.731990Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:16:54.731990Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Peri-LN: Revisiting Normalization Layer in the Transformer Architecture","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Beomjun Kim, Byeongchan Lee, Cheonbok Park, Dongyoon Han, Jeonghoon Kim, Jinwoo Shin, Kang Min Yoo, Seongjin Shin, Taehwan Yoo, Yeontaek Oh","submitted_at":"2025-02-04T21:29:47Z","abstract_excerpt":"Selecting a layer normalization (LN) strategy that stabilizes training and speeds convergence in Transformers remains difficult, even for today's large language models (LLM). We present a comprehensive analytical foundation for understanding how different LN strategies influence training dynamics in large-scale Transformers. Until recently, Pre-LN and Post-LN have long dominated practices despite their limitations in large-scale training. However, several open-source models have recently begun silently adopting a third strategy without much explanation. This strategy places normalization layer"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.02732","kind":"arxiv","version":3},"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/2502.02732/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":"2502.02732","created_at":"2026-07-05T11:16:54.732048+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.02732v3","created_at":"2026-07-05T11:16:54.732048+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.02732","created_at":"2026-07-05T11:16:54.732048+00:00"},{"alias_kind":"pith_short_12","alias_value":"PHQ7GOGAAQRW","created_at":"2026-07-05T11:16:54.732048+00:00"},{"alias_kind":"pith_short_16","alias_value":"PHQ7GOGAAQRW2DV6","created_at":"2026-07-05T11:16:54.732048+00:00"},{"alias_kind":"pith_short_8","alias_value":"PHQ7GOGA","created_at":"2026-07-05T11:16:54.732048+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25971","citing_title":"Improving Neural Network Training by Decoupling the Magnitude and Direction of Weight Vectors","ref_index":98,"is_internal_anchor":false},{"citing_arxiv_id":"2606.19491","citing_title":"Algebraic Dead Directions in LayerNorm Transformers: A Forward-Pass-Only Diagnostic at LLM Scale","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05957","citing_title":"Dead Directions: Geometric Singular Learning","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2602.08064","citing_title":"SiameseNorm: Breaking the Barrier to Reconciling Pre/Post-Norm","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23434","citing_title":"When Does Removing LayerNorm Help? Activation Bounding as a Regime-Dependent Implicit Regularizer","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15259","citing_title":"Stability and Generalization in Looped Transformers","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PHQ7GOGAAQRW2DV6AVSM2XPZ5R","json":"https://pith.science/pith/PHQ7GOGAAQRW2DV6AVSM2XPZ5R.json","graph_json":"https://pith.science/api/pith-number/PHQ7GOGAAQRW2DV6AVSM2XPZ5R/graph.json","events_json":"https://pith.science/api/pith-number/PHQ7GOGAAQRW2DV6AVSM2XPZ5R/events.json","paper":"https://pith.science/paper/PHQ7GOGA"},"agent_actions":{"view_html":"https://pith.science/pith/PHQ7GOGAAQRW2DV6AVSM2XPZ5R","download_json":"https://pith.science/pith/PHQ7GOGAAQRW2DV6AVSM2XPZ5R.json","view_paper":"https://pith.science/paper/PHQ7GOGA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.02732&json=true","fetch_graph":"https://pith.science/api/pith-number/PHQ7GOGAAQRW2DV6AVSM2XPZ5R/graph.json","fetch_events":"https://pith.science/api/pith-number/PHQ7GOGAAQRW2DV6AVSM2XPZ5R/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PHQ7GOGAAQRW2DV6AVSM2XPZ5R/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PHQ7GOGAAQRW2DV6AVSM2XPZ5R/action/storage_attestation","attest_author":"https://pith.science/pith/PHQ7GOGAAQRW2DV6AVSM2XPZ5R/action/author_attestation","sign_citation":"https://pith.science/pith/PHQ7GOGAAQRW2DV6AVSM2XPZ5R/action/citation_signature","submit_replication":"https://pith.science/pith/PHQ7GOGAAQRW2DV6AVSM2XPZ5R/action/replication_record"}},"created_at":"2026-07-05T11:16:54.732048+00:00","updated_at":"2026-07-05T11:16:54.732048+00:00"}