{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:T3YTLRY7KMENON7GG37S6IEDFH","short_pith_number":"pith:T3YTLRY7","schema_version":"1.0","canonical_sha256":"9ef135c71f5308d737e636ff2f208329d5970a3baf0855c80e934291e027e335","source":{"kind":"arxiv","id":"2507.16676","version":1},"attestation_state":"computed","paper":{"title":"Custom Algorithm-based Fault Tolerance for Attention Layers in Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AR"],"primary_cat":"cs.LG","authors_text":"Giorgos Dimitrakopoulos, Kosmas Alexandridis, Vasileios Titopoulos","submitted_at":"2025-07-22T15:11:13Z","abstract_excerpt":"Transformers and large language models (LLMs), powered by the attention mechanism, have transformed numerous AI applications, driving the need for specialized hardware accelerators. A major challenge in these accelerators is efficiently detecting errors caused by random hardware faults. Traditional algorithm-based fault tolerance (ABFT) techniques verify individual matrix multiplications but fall short in handling the full attention mechanism, particularly due to intermediate softmax normalization. This work proposes Flash-ABFT, a novel method that computes an online checksum across the entire"},"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":"2507.16676","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-07-22T15:11:13Z","cross_cats_sorted":["cs.AR"],"title_canon_sha256":"b9d4d7ba5d2ebe0b5692ce457465ff65588f95f42445cf540768f71a947e3ef7","abstract_canon_sha256":"5d31d34226a036c424316be5c65e3dc02eb1fc5810131b1522aca0be106275a2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:41:28.964234Z","signature_b64":"wtbrdyvwOtAMkYJjp434/jq8iy4nkuWx8g8icRwaFsZln2KVjOa4wfaaDxMP6wk44xHyotB0FjbQ6qhTcb7gAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9ef135c71f5308d737e636ff2f208329d5970a3baf0855c80e934291e027e335","last_reissued_at":"2026-07-05T11:41:28.963772Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:41:28.963772Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Custom Algorithm-based Fault Tolerance for Attention Layers in Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AR"],"primary_cat":"cs.LG","authors_text":"Giorgos Dimitrakopoulos, Kosmas Alexandridis, Vasileios Titopoulos","submitted_at":"2025-07-22T15:11:13Z","abstract_excerpt":"Transformers and large language models (LLMs), powered by the attention mechanism, have transformed numerous AI applications, driving the need for specialized hardware accelerators. A major challenge in these accelerators is efficiently detecting errors caused by random hardware faults. Traditional algorithm-based fault tolerance (ABFT) techniques verify individual matrix multiplications but fall short in handling the full attention mechanism, particularly due to intermediate softmax normalization. This work proposes Flash-ABFT, a novel method that computes an online checksum across the entire"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.16676","kind":"arxiv","version":1},"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/2507.16676/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":"2507.16676","created_at":"2026-07-05T11:41:28.963828+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.16676v1","created_at":"2026-07-05T11:41:28.963828+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.16676","created_at":"2026-07-05T11:41:28.963828+00:00"},{"alias_kind":"pith_short_12","alias_value":"T3YTLRY7KMEN","created_at":"2026-07-05T11:41:28.963828+00:00"},{"alias_kind":"pith_short_16","alias_value":"T3YTLRY7KMENON7G","created_at":"2026-07-05T11:41:28.963828+00:00"},{"alias_kind":"pith_short_8","alias_value":"T3YTLRY7","created_at":"2026-07-05T11:41:28.963828+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.02430","citing_title":"Not All Errors Are Equal: A Systematic Study of Error Propagation in Large Language Model Inference","ref_index":80,"is_internal_anchor":false},{"citing_arxiv_id":"2606.27934","citing_title":"Self-Verifying Measurement Records: Hash-Linked Evidence Graphs for Hardware Benchmarking","ref_index":45,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T3YTLRY7KMENON7GG37S6IEDFH","json":"https://pith.science/pith/T3YTLRY7KMENON7GG37S6IEDFH.json","graph_json":"https://pith.science/api/pith-number/T3YTLRY7KMENON7GG37S6IEDFH/graph.json","events_json":"https://pith.science/api/pith-number/T3YTLRY7KMENON7GG37S6IEDFH/events.json","paper":"https://pith.science/paper/T3YTLRY7"},"agent_actions":{"view_html":"https://pith.science/pith/T3YTLRY7KMENON7GG37S6IEDFH","download_json":"https://pith.science/pith/T3YTLRY7KMENON7GG37S6IEDFH.json","view_paper":"https://pith.science/paper/T3YTLRY7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.16676&json=true","fetch_graph":"https://pith.science/api/pith-number/T3YTLRY7KMENON7GG37S6IEDFH/graph.json","fetch_events":"https://pith.science/api/pith-number/T3YTLRY7KMENON7GG37S6IEDFH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T3YTLRY7KMENON7GG37S6IEDFH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T3YTLRY7KMENON7GG37S6IEDFH/action/storage_attestation","attest_author":"https://pith.science/pith/T3YTLRY7KMENON7GG37S6IEDFH/action/author_attestation","sign_citation":"https://pith.science/pith/T3YTLRY7KMENON7GG37S6IEDFH/action/citation_signature","submit_replication":"https://pith.science/pith/T3YTLRY7KMENON7GG37S6IEDFH/action/replication_record"}},"created_at":"2026-07-05T11:41:28.963828+00:00","updated_at":"2026-07-05T11:41:28.963828+00:00"}