{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HTDI2FY7TGIY4JUVGQOBQPCSCP","short_pith_number":"pith:HTDI2FY7","schema_version":"1.0","canonical_sha256":"3cc68d171f99918e2695341c183c5213d50f9aeb7244fbc0a865f971fad7dd10","source":{"kind":"arxiv","id":"2601.22580","version":2},"attestation_state":"computed","paper":{"title":"SpanNorm: Reconciling Training Stability and Performance in Deep Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bei Li, Chao Wang, Jiaqi Zhang, Jingang Wang, Linkun Lyu, Peng Pei, Tong Xiao, Xin Chen, Xinyu Liu, Xunliang Cai, Yuchun Fan","submitted_at":"2026-01-30T05:21:57Z","abstract_excerpt":"The success of Large Language Models (LLMs) hinges on the stable training of deep Transformer architectures. A critical design choice is the placement of normalization layers, leading to a fundamental trade-off: the ``PreNorm'' architecture ensures training stability at the cost of potential performance degradation in deep models, while the ``PostNorm'' architecture offers strong performance but suffers from severe training instability. In this work, we propose SpanNorm, a novel technique designed to resolve this dilemma by integrating the strengths of both paradigms. Structurally, SpanNorm es"},"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":"2601.22580","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-01-30T05:21:57Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"bb1cb96f374fafa4d11f48d916fac452a7f2bfd25ae4ebcb1a2722dd1055e08b","abstract_canon_sha256":"0166ef561633e6c530ab93cf86303531d193319f8f5d530af5fc8eccf394609a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-05T01:15:19.556084Z","signature_b64":"11zMnb/DWwgCOcVcd5qxVUuD24VkFhr1MgBEA8sHtAcUHMsfpbiU7swtMiOsmmVBNvXLts6wDlo0EyOMlGpNCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3cc68d171f99918e2695341c183c5213d50f9aeb7244fbc0a865f971fad7dd10","last_reissued_at":"2026-06-05T01:15:19.555586Z","signature_status":"signed_v1","first_computed_at":"2026-06-05T01:15:19.555586Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SpanNorm: Reconciling Training Stability and Performance in Deep Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bei Li, Chao Wang, Jiaqi Zhang, Jingang Wang, Linkun Lyu, Peng Pei, Tong Xiao, Xin Chen, Xinyu Liu, Xunliang Cai, Yuchun Fan","submitted_at":"2026-01-30T05:21:57Z","abstract_excerpt":"The success of Large Language Models (LLMs) hinges on the stable training of deep Transformer architectures. A critical design choice is the placement of normalization layers, leading to a fundamental trade-off: the ``PreNorm'' architecture ensures training stability at the cost of potential performance degradation in deep models, while the ``PostNorm'' architecture offers strong performance but suffers from severe training instability. In this work, we propose SpanNorm, a novel technique designed to resolve this dilemma by integrating the strengths of both paradigms. Structurally, SpanNorm es"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2601.22580","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/2601.22580/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":"2601.22580","created_at":"2026-06-05T01:15:19.555648+00:00"},{"alias_kind":"arxiv_version","alias_value":"2601.22580v2","created_at":"2026-06-05T01:15:19.555648+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.22580","created_at":"2026-06-05T01:15:19.555648+00:00"},{"alias_kind":"pith_short_12","alias_value":"HTDI2FY7TGIY","created_at":"2026-06-05T01:15:19.555648+00:00"},{"alias_kind":"pith_short_16","alias_value":"HTDI2FY7TGIY4JUV","created_at":"2026-06-05T01:15:19.555648+00:00"},{"alias_kind":"pith_short_8","alias_value":"HTDI2FY7","created_at":"2026-06-05T01:15:19.555648+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HTDI2FY7TGIY4JUVGQOBQPCSCP","json":"https://pith.science/pith/HTDI2FY7TGIY4JUVGQOBQPCSCP.json","graph_json":"https://pith.science/api/pith-number/HTDI2FY7TGIY4JUVGQOBQPCSCP/graph.json","events_json":"https://pith.science/api/pith-number/HTDI2FY7TGIY4JUVGQOBQPCSCP/events.json","paper":"https://pith.science/paper/HTDI2FY7"},"agent_actions":{"view_html":"https://pith.science/pith/HTDI2FY7TGIY4JUVGQOBQPCSCP","download_json":"https://pith.science/pith/HTDI2FY7TGIY4JUVGQOBQPCSCP.json","view_paper":"https://pith.science/paper/HTDI2FY7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2601.22580&json=true","fetch_graph":"https://pith.science/api/pith-number/HTDI2FY7TGIY4JUVGQOBQPCSCP/graph.json","fetch_events":"https://pith.science/api/pith-number/HTDI2FY7TGIY4JUVGQOBQPCSCP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HTDI2FY7TGIY4JUVGQOBQPCSCP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HTDI2FY7TGIY4JUVGQOBQPCSCP/action/storage_attestation","attest_author":"https://pith.science/pith/HTDI2FY7TGIY4JUVGQOBQPCSCP/action/author_attestation","sign_citation":"https://pith.science/pith/HTDI2FY7TGIY4JUVGQOBQPCSCP/action/citation_signature","submit_replication":"https://pith.science/pith/HTDI2FY7TGIY4JUVGQOBQPCSCP/action/replication_record"}},"created_at":"2026-06-05T01:15:19.555648+00:00","updated_at":"2026-06-05T01:15:19.555648+00:00"}