{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:M5RS33JCA3HRS4V7PNWWJCGPRE","short_pith_number":"pith:M5RS33JC","schema_version":"1.0","canonical_sha256":"67632ded2206cf1972bf7b6d6488cf89298b0156193769865b0a5d318476d934","source":{"kind":"arxiv","id":"2510.21770","version":2},"attestation_state":"computed","paper":{"title":"Numerical Fragility in Transformers: A Layer-wise Theory for Risk Estimation and Selective Stabilization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.NA","math.NA"],"primary_cat":"cs.LG","authors_text":"Jinwoo Baek","submitted_at":"2025-10-17T01:03:02Z","abstract_excerpt":"Low-precision execution can induce substantial forward discrepancies in Transformers even for fixed weights and input, yet these discrepancies are usually monitored only at the output and lack a layer-wise theoretical account. We develop a first-order decomposition of output mismatch into layer-local attention, LayerNorm, and residual-transport terms, and derive from it a practical causal risk estimator and a budgeted controller, Bound-Guided Selective Stabilization (BGSS). Controlled sweeps verify the predicted local sign, monotonicity, and transport structure. On GPT-2, the transport-aware c"},"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":"2510.21770","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-10-17T01:03:02Z","cross_cats_sorted":["cs.NA","math.NA"],"title_canon_sha256":"ba9aeaecb0020095c3bb5751cd8aa0773f6c132b8ab7ad69f3ef111c554337dd","abstract_canon_sha256":"4dd1c902e9eb927834d71301ad8e1af19923991e1e13b447f6a91a06309ac8ba"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-27T01:19:51.447449Z","signature_b64":"Ff583Z/oMBGuFA2aRfThA+PVcul3zokrrSJNElZrG8c5DZ2OBiCSdv5KK7QVvIbliR1G+OmfIKunu/wjB7/yBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"67632ded2206cf1972bf7b6d6488cf89298b0156193769865b0a5d318476d934","last_reissued_at":"2026-07-27T01:19:51.446630Z","signature_status":"signed_v1","first_computed_at":"2026-07-27T01:19:51.446630Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Numerical Fragility in Transformers: A Layer-wise Theory for Risk Estimation and Selective Stabilization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.NA","math.NA"],"primary_cat":"cs.LG","authors_text":"Jinwoo Baek","submitted_at":"2025-10-17T01:03:02Z","abstract_excerpt":"Low-precision execution can induce substantial forward discrepancies in Transformers even for fixed weights and input, yet these discrepancies are usually monitored only at the output and lack a layer-wise theoretical account. We develop a first-order decomposition of output mismatch into layer-local attention, LayerNorm, and residual-transport terms, and derive from it a practical causal risk estimator and a budgeted controller, Bound-Guided Selective Stabilization (BGSS). Controlled sweeps verify the predicted local sign, monotonicity, and transport structure. On GPT-2, the transport-aware c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.21770","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/2510.21770/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":"2510.21770","created_at":"2026-07-27T01:19:51.447030+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.21770v2","created_at":"2026-07-27T01:19:51.447030+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.21770","created_at":"2026-07-27T01:19:51.447030+00:00"},{"alias_kind":"pith_short_12","alias_value":"M5RS33JCA3HR","created_at":"2026-07-27T01:19:51.447030+00:00"},{"alias_kind":"pith_short_16","alias_value":"M5RS33JCA3HRS4V7","created_at":"2026-07-27T01:19:51.447030+00:00"},{"alias_kind":"pith_short_8","alias_value":"M5RS33JC","created_at":"2026-07-27T01:19:51.447030+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/M5RS33JCA3HRS4V7PNWWJCGPRE","json":"https://pith.science/pith/M5RS33JCA3HRS4V7PNWWJCGPRE.json","graph_json":"https://pith.science/api/pith-number/M5RS33JCA3HRS4V7PNWWJCGPRE/graph.json","events_json":"https://pith.science/api/pith-number/M5RS33JCA3HRS4V7PNWWJCGPRE/events.json","paper":"https://pith.science/paper/M5RS33JC"},"agent_actions":{"view_html":"https://pith.science/pith/M5RS33JCA3HRS4V7PNWWJCGPRE","download_json":"https://pith.science/pith/M5RS33JCA3HRS4V7PNWWJCGPRE.json","view_paper":"https://pith.science/paper/M5RS33JC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.21770&json=true","fetch_graph":"https://pith.science/api/pith-number/M5RS33JCA3HRS4V7PNWWJCGPRE/graph.json","fetch_events":"https://pith.science/api/pith-number/M5RS33JCA3HRS4V7PNWWJCGPRE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/M5RS33JCA3HRS4V7PNWWJCGPRE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/M5RS33JCA3HRS4V7PNWWJCGPRE/action/storage_attestation","attest_author":"https://pith.science/pith/M5RS33JCA3HRS4V7PNWWJCGPRE/action/author_attestation","sign_citation":"https://pith.science/pith/M5RS33JCA3HRS4V7PNWWJCGPRE/action/citation_signature","submit_replication":"https://pith.science/pith/M5RS33JCA3HRS4V7PNWWJCGPRE/action/replication_record"}},"created_at":"2026-07-27T01:19:51.447030+00:00","updated_at":"2026-07-27T01:19:51.447030+00:00"}