{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:MNVLB3NCEIFLXJYEQTU4XPUT7W","short_pith_number":"pith:MNVLB3NC","schema_version":"1.0","canonical_sha256":"636ab0eda2220abba70484e9cbbe93fda78270dac698dadbd4e9aa4354d32bd1","source":{"kind":"arxiv","id":"2504.12991","version":2},"attestation_state":"computed","paper":{"title":"A Theoretical Framework for OOD Robustness in Transformers using Gevrey Classes","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Fu-Chieh Chang, Pei-Yuan Wu, Yu Wang","submitted_at":"2025-04-17T14:59:29Z","abstract_excerpt":"We study the robustness of Transformer language models under semantic out-of-distribution (OOD) shifts, where training and test data lie in disjoint latent spaces. Using Wasserstein-1 distance and Gevrey-class smoothness, we derive sub-exponential upper bounds on prediction error. Our theoretical framework explains how smoothness governs generalization under distributional drift. We validate these findings through controlled experiments on arithmetic and Chain-of-Thought tasks with latent permutations and scalings. Results show empirical degradation aligns with our bounds, highlighting the geo"},"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":"2504.12991","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-17T14:59:29Z","cross_cats_sorted":[],"title_canon_sha256":"47e92a317781d76d83d2b3081db56ef551662de7462c29ee7a2cbf5353d71e78","abstract_canon_sha256":"46f19422c7197c15dda36af298affc6cb84fad98fc68a7c72f115ba5ac90edf1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:12:30.915506Z","signature_b64":"w6numKy+8jVEvyyjieo7S1ZB6ycq9ullBrE6YtHYAebnWJ2TVaRwpfNcZ3uWVQDHMXN3uHg7E6gAFQwVnwtuBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"636ab0eda2220abba70484e9cbbe93fda78270dac698dadbd4e9aa4354d32bd1","last_reissued_at":"2026-07-05T11:12:30.914988Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:12:30.914988Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Theoretical Framework for OOD Robustness in Transformers using Gevrey Classes","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Fu-Chieh Chang, Pei-Yuan Wu, Yu Wang","submitted_at":"2025-04-17T14:59:29Z","abstract_excerpt":"We study the robustness of Transformer language models under semantic out-of-distribution (OOD) shifts, where training and test data lie in disjoint latent spaces. Using Wasserstein-1 distance and Gevrey-class smoothness, we derive sub-exponential upper bounds on prediction error. Our theoretical framework explains how smoothness governs generalization under distributional drift. We validate these findings through controlled experiments on arithmetic and Chain-of-Thought tasks with latent permutations and scalings. Results show empirical degradation aligns with our bounds, highlighting the geo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.12991","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/2504.12991/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":"2504.12991","created_at":"2026-07-05T11:12:30.915050+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.12991v2","created_at":"2026-07-05T11:12:30.915050+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.12991","created_at":"2026-07-05T11:12:30.915050+00:00"},{"alias_kind":"pith_short_12","alias_value":"MNVLB3NCEIFL","created_at":"2026-07-05T11:12:30.915050+00:00"},{"alias_kind":"pith_short_16","alias_value":"MNVLB3NCEIFLXJYE","created_at":"2026-07-05T11:12:30.915050+00:00"},{"alias_kind":"pith_short_8","alias_value":"MNVLB3NC","created_at":"2026-07-05T11:12:30.915050+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2508.01191","citing_title":"Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MNVLB3NCEIFLXJYEQTU4XPUT7W","json":"https://pith.science/pith/MNVLB3NCEIFLXJYEQTU4XPUT7W.json","graph_json":"https://pith.science/api/pith-number/MNVLB3NCEIFLXJYEQTU4XPUT7W/graph.json","events_json":"https://pith.science/api/pith-number/MNVLB3NCEIFLXJYEQTU4XPUT7W/events.json","paper":"https://pith.science/paper/MNVLB3NC"},"agent_actions":{"view_html":"https://pith.science/pith/MNVLB3NCEIFLXJYEQTU4XPUT7W","download_json":"https://pith.science/pith/MNVLB3NCEIFLXJYEQTU4XPUT7W.json","view_paper":"https://pith.science/paper/MNVLB3NC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.12991&json=true","fetch_graph":"https://pith.science/api/pith-number/MNVLB3NCEIFLXJYEQTU4XPUT7W/graph.json","fetch_events":"https://pith.science/api/pith-number/MNVLB3NCEIFLXJYEQTU4XPUT7W/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MNVLB3NCEIFLXJYEQTU4XPUT7W/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MNVLB3NCEIFLXJYEQTU4XPUT7W/action/storage_attestation","attest_author":"https://pith.science/pith/MNVLB3NCEIFLXJYEQTU4XPUT7W/action/author_attestation","sign_citation":"https://pith.science/pith/MNVLB3NCEIFLXJYEQTU4XPUT7W/action/citation_signature","submit_replication":"https://pith.science/pith/MNVLB3NCEIFLXJYEQTU4XPUT7W/action/replication_record"}},"created_at":"2026-07-05T11:12:30.915050+00:00","updated_at":"2026-07-05T11:12:30.915050+00:00"}