{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:MEFZ47YPYK7CJ7SUKJUBDQ75NS","short_pith_number":"pith:MEFZ47YP","canonical_record":{"source":{"id":"2604.28118","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-04-30T17:07:11Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"51cece6c498dc2d9af2b522f877e6dd9698e53e74233fe706d0fa66424f125b3","abstract_canon_sha256":"821545c2ea291801ec636cb20017331dd9a91be57be4b261ceec495a923e3c1f"},"schema_version":"1.0"},"canonical_sha256":"610b9e7f0fc2be24fe54526811c3fd6cabb2efdb5e57da660681178afb9b51cd","source":{"kind":"arxiv","id":"2604.28118","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.28118","created_at":"2026-06-26T01:15:19Z"},{"alias_kind":"arxiv_version","alias_value":"2604.28118v2","created_at":"2026-06-26T01:15:19Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.28118","created_at":"2026-06-26T01:15:19Z"},{"alias_kind":"pith_short_12","alias_value":"MEFZ47YPYK7C","created_at":"2026-06-26T01:15:19Z"},{"alias_kind":"pith_short_16","alias_value":"MEFZ47YPYK7CJ7SU","created_at":"2026-06-26T01:15:19Z"},{"alias_kind":"pith_short_8","alias_value":"MEFZ47YP","created_at":"2026-06-26T01:15:19Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:MEFZ47YPYK7CJ7SUKJUBDQ75NS","target":"record","payload":{"canonical_record":{"source":{"id":"2604.28118","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-04-30T17:07:11Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"51cece6c498dc2d9af2b522f877e6dd9698e53e74233fe706d0fa66424f125b3","abstract_canon_sha256":"821545c2ea291801ec636cb20017331dd9a91be57be4b261ceec495a923e3c1f"},"schema_version":"1.0"},"canonical_sha256":"610b9e7f0fc2be24fe54526811c3fd6cabb2efdb5e57da660681178afb9b51cd","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-26T01:15:19.009964Z","signature_b64":"d9nSISqHVV3mZhmZCXlwEIpOeGhTQ7Q9iJovPm056BW569VL7r9HHOOsW/CfrRznLl0FjxSdjrVae+L1st7cAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"610b9e7f0fc2be24fe54526811c3fd6cabb2efdb5e57da660681178afb9b51cd","last_reissued_at":"2026-06-26T01:15:19.009534Z","signature_status":"signed_v1","first_computed_at":"2026-06-26T01:15:19.009534Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.28118","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-26T01:15:19Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"soL0PbK5dhdEtns68h24DQbES9Y/Er+kohk6bNGw+Xy47CRwTLmVPvfu+hjkrUuEMFR0hYGIBDc3DjjREjMTCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T03:19:51.949581Z"},"content_sha256":"42ee59fbae50ccaf9ee10d31184ebfe32f1f32cc6046390baf546b19a9d57f3b","schema_version":"1.0","event_id":"sha256:42ee59fbae50ccaf9ee10d31184ebfe32f1f32cc6046390baf546b19a9d57f3b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:MEFZ47YPYK7CJ7SUKJUBDQ75NS","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Hierarchical Fault Detection and Diagnosis for Transformer Architectures","license":"http://creativecommons.org/licenses/by/4.0/","headline":"DEFault++ detects whether a fault exists in a transformer model, classifies it into one of 12 specific categories, and identifies its root cause among up to 45 mechanisms by analyzing runtime behaviors through an architecture-derived graph.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.SE","authors_text":"Mohammad Masudur Rahman, Saurabh Singh Rajput, Sigma Jahan, Tushar Sharma","submitted_at":"2026-04-30T17:07:11Z","abstract_excerpt":"Transformers now underpin critical AI systems across industry and research. Yet their faults can silently alter model behavior without runtime errors, and existing techniques offer little support for tracing these failures to their component and root cause. Such faults evade detection because loss and numerical values stay normal, and the visible symptom rarely identifies the component responsible. We present DEFault++, a hierarchical learning-based technique that first detects a fault, then identifies the affected component, and finally the cause within it, helping developers effectively debu"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On DEFault-bench, DEFault++ exceeds an AUROC of 0.96 for detection and a Macro-F1 of 0.85 for both categorization and root-cause diagnosis on encoder and decoder architectures. In a developer study with 21 practitioners, the accuracy of choosing correct repair actions increased from 57.1% without support to 83.3% when using DEFault++.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The faults generated by the DEForm mutation technique accurately represent the kinds of faults that occur in real deployed transformer models, and the Fault Propagation Graph derived from the architecture fully captures how those faults affect observable behavior.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DEFault++ delivers automated hierarchical fault detection, categorization into 12 transformer-specific types, and root-cause diagnosis among 45 mechanisms on a new benchmark of 3,739 mutated instances, with AUROC >0.96 and Macro-F1 0.85, plus improved developer repair accuracy in a user study.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"DEFault++ detects whether a fault exists in a transformer model, classifies it into one of 12 specific categories, and identifies its root cause among up to 45 mechanisms by analyzing runtime behaviors through an architecture-derived graph.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"74b83fc509150acb52609245e25cf606d384188cca259d1f4fc71dfc4fc27bc2"},"source":{"id":"2604.28118","kind":"arxiv","version":2},"verdict":{"id":"223c4a4d-aa2a-491d-8bb6-b22fac962e50","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T05:28:19.157068Z","strongest_claim":"On DEFault-bench, DEFault++ exceeds an AUROC of 0.96 for detection and a Macro-F1 of 0.85 for both categorization and root-cause diagnosis on encoder and decoder architectures. In a developer study with 21 practitioners, the accuracy of choosing correct repair actions increased from 57.1% without support to 83.3% when using DEFault++.","one_line_summary":"DEFault++ delivers automated hierarchical fault detection, categorization into 12 transformer-specific types, and root-cause diagnosis among 45 mechanisms on a new benchmark of 3,739 mutated instances, with AUROC >0.96 and Macro-F1 0.85, plus improved developer repair accuracy in a user study.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The faults generated by the DEForm mutation technique accurately represent the kinds of faults that occur in real deployed transformer models, and the Fault Propagation Graph derived from the architecture fully captures how those faults affect observable behavior.","pith_extraction_headline":"DEFault++ detects whether a fault exists in a transformer model, classifies it into one of 12 specific categories, and identifies its root cause among up to 45 mechanisms by analyzing runtime behaviors through an architecture-derived graph."},"integrity":{"clean":false,"summary":{"advisory":2,"critical":1,"by_detector":{"doi_compliance":{"total":3,"advisory":2,"critical":1,"informational":0}},"informational":0},"endpoint":"/pith/2604.28118/integrity.json","findings":[{"note":"Identifier '10.1109/ase.2019.000' is syntactically valid but the DOI registry (doi.org) returned 404, and Crossref / OpenAlex / internal corpus also have no record. The cited work could not be located through any authoritative source.","detector":"doi_compliance","severity":"critical","ref_index":40,"audited_at":"2026-05-19T18:36:39.919630Z","detected_doi":"10.1109/ase.2019.000","finding_type":"unresolvable_identifier","verdict_class":"cross_source","detected_arxiv_id":null},{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1145/3180155.3180220.Ashish) was visible in the surrounding text but could not be confirmed against doi.org as printed.","detector":"doi_compliance","severity":"advisory","ref_index":34,"audited_at":"2026-05-19T18:36:39.919630Z","detected_doi":"10.1145/3180155.3180220.Ashish","finding_type":"recoverable_identifier","verdict_class":"incontrovertible","detected_arxiv_id":null},{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1109/TSE.2025.3543187.Ketai) was visible in the surrounding text but could not be confirmed against doi.org as printed.","detector":"doi_compliance","severity":"advisory","ref_index":27,"audited_at":"2026-05-19T18:36:39.919630Z","detected_doi":"10.1109/TSE.2025.3543187.Ketai","finding_type":"recoverable_identifier","verdict_class":"incontrovertible","detected_arxiv_id":null}],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T20:40:38.128942Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T18:36:39.919630Z","status":"completed","version":"1.0.0","findings_count":3}],"snapshot_sha256":"9d7c299ad0f4190d94e5474087717ec45a38a25b17e28865ce4b20545c27be55"},"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"},"verdict_id":"223c4a4d-aa2a-491d-8bb6-b22fac962e50"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-26T01:15:19Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"BJ1IZRINAP57vbgKWJi/wZrV3Ag0RSiF6FOfQm3t5jMRZ/psRy4qTRUGuRdwKlNVdSlHKXnQdhGzbx8n4G2ZAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T03:19:51.950235Z"},"content_sha256":"f4a2dc9537a305baadf7f7668c254be1e297ec8776eb07d7c7df1ef0ec046296","schema_version":"1.0","event_id":"sha256:f4a2dc9537a305baadf7f7668c254be1e297ec8776eb07d7c7df1ef0ec046296"},{"event_type":"integrity_finding","subject_pith_number":"pith:2026:MEFZ47YPYK7CJ7SUKJUBDQ75NS","target":"integrity","payload":{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1145/3180155.3180220.Ashish) was visible in the surrounding text but could not be confirmed against doi.org as printed.","snippet":"ACM. doi: 10.1145/31 80155.3180220. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. InAdvances in Neural Information Processing Systems,","arxiv_id":"2604.28118","detector":"doi_compliance","evidence":{"ref_index":34,"verdict_class":"incontrovertible","resolved_title":null,"printed_excerpt":"ACM. doi: 10.1145/31 80155.3180220. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. InAdvances in Neural Information Processing Systems,","reconstructed_doi":"10.1145/3180155.3180220.Ashish"},"severity":"advisory","ref_index":34,"audited_at":"2026-05-19T18:36:39.919630Z","event_type":"pith.integrity.v1","detected_doi":"10.1145/3180155.3180220.Ashish","detector_url":"https://pith.science/pith-integrity-protocol#doi_compliance","external_url":null,"finding_type":"recoverable_identifier","evidence_hash":"c1545491f1f82d7c7ce79c83af8acbfeac47201843a3612fbbe9056a35b1c244","paper_version":1,"verdict_class":"incontrovertible","resolved_title":null,"detector_version":"1.0.0","detected_arxiv_id":null,"integrity_event_id":2514,"payload_sha256":"cf3bb1f4277bef1e4914c5aea65bea1906cbeb17eeeda050b78c16c382866e05","signature_b64":"jNV+EjCbvtCL2P2drbDFtAHPesyZ2uFf7hl88OtiNalpTMn/0oVmCiR2Bzaleojs5dbG5MBoFiM7mUqf7+XnDQ==","signing_key_id":"pith-v1-2026-05"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-19T18:37:17Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"SRDQFU0q061xexTJzvKs2smPjwJ0AtJRyV7gPUfCFGRcm3WKT0DZ/jZam4kdC2Gj0CleWwuvWGkXuCzhefw6Ag==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T03:19:51.951889Z"},"content_sha256":"f4c8bed3a4c4fdedd212eacdafbfc697a197320d3cfc576321f3da1c0b55a2d6","schema_version":"1.0","event_id":"sha256:f4c8bed3a4c4fdedd212eacdafbfc697a197320d3cfc576321f3da1c0b55a2d6"},{"event_type":"integrity_finding","subject_pith_number":"pith:2026:MEFZ47YPYK7CJ7SUKJUBDQ75NS","target":"integrity","payload":{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1109/TSE.2025.3543187.Ketai) was visible in the surrounding text but could not be confirmed against doi.org as printed.","snippet":"doi: 10.1109/TSE.2025.3 543187. Ketai Qiu, Niccolò Puccinelli, Matteo Ciniselli, and Luca Di Grazia. From today’s code to tomor- row’s symphony: The ai transformation of developer’s routine by 2030.ACM Transactions on Software Engineering a","arxiv_id":"2604.28118","detector":"doi_compliance","evidence":{"ref_index":27,"verdict_class":"incontrovertible","resolved_title":null,"printed_excerpt":"doi: 10.1109/TSE.2025.3 543187. Ketai Qiu, Niccolò Puccinelli, Matteo Ciniselli, and Luca Di Grazia. From today’s code to tomor- row’s symphony: The ai transformation of developer’s routine by 2030.ACM Transactions on Software Engineering a","reconstructed_doi":"10.1109/TSE.2025.3543187.Ketai"},"severity":"advisory","ref_index":27,"audited_at":"2026-05-19T18:36:39.919630Z","event_type":"pith.integrity.v1","detected_doi":"10.1109/TSE.2025.3543187.Ketai","detector_url":"https://pith.science/pith-integrity-protocol#doi_compliance","external_url":null,"finding_type":"recoverable_identifier","evidence_hash":"231cecce31bb52f055aae167190fdfe4bd6f157427c1f7b85ebab01d2b0eaeba","paper_version":1,"verdict_class":"incontrovertible","resolved_title":null,"detector_version":"1.0.0","detected_arxiv_id":null,"integrity_event_id":2513,"payload_sha256":"bec61fa71ef1236b71eaf745a53a9059f39d52937e90acc396bf8e4a69ffab7e","signature_b64":"Pcj5C7xMTOc9cLRJ5TQ9oJYhVMdR7kr/4Pw9ZhcVjsjFhmhO54qrpY46rsW2WLFVVvzRF8ezEPqCn2K8H/F8CQ==","signing_key_id":"pith-v1-2026-05"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-19T18:37:17Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"i915K1naWZkyXamG436FNpwKt6zcVkbyxgzlA1dtUFzZlY7HVW+4dwQFeCUGOjETNsRwfDSlSexZFm/1WtF9Dg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T03:19:51.952187Z"},"content_sha256":"4756a26d17ab3411eb7e5d1ada67e75428c0f930c07827e0e67fc26c96092fa3","schema_version":"1.0","event_id":"sha256:4756a26d17ab3411eb7e5d1ada67e75428c0f930c07827e0e67fc26c96092fa3"},{"event_type":"integrity_finding","subject_pith_number":"pith:2026:MEFZ47YPYK7CJ7SUKJUBDQ75NS","target":"integrity","payload":{"note":"Identifier '10.1109/ase.2019.000' is syntactically valid but the DOI registry (doi.org) returned 404, and Crossref / OpenAlex / internal corpus also have no record. The cited work could not be located through any authoritative source.","snippet":"IEEE. doi: 10.1109/ASE.2019.000","arxiv_id":"2604.28118","detector":"doi_compliance","evidence":{"doi":"10.1109/ase.2019.000","arxiv_id":null,"ref_index":40,"raw_excerpt":"IEEE. doi: 10.1109/ASE.2019.000","verdict_class":"cross_source","checked_sources":["crossref_by_doi","openalex_by_doi","doi_org_head"]},"severity":"critical","ref_index":40,"audited_at":"2026-05-19T18:36:39.919630Z","event_type":"pith.integrity.v1","detected_doi":"10.1109/ase.2019.000","detector_url":"https://pith.science/pith-integrity-protocol#doi_compliance","external_url":null,"finding_type":"unresolvable_identifier","evidence_hash":"c94a8c939c56548a328ccafa5a80672d310867292e07ef74fab3819fec1bfe13","paper_version":1,"verdict_class":"cross_source","resolved_title":null,"detector_version":"1.0.0","detected_arxiv_id":null,"integrity_event_id":2512,"payload_sha256":"e1d9fdb4b382b6d094638653eb362168cb809d64e95acdb7e9ba17d42afe7548","signature_b64":"f32eDyFh/5+obeHb4Ryeoypbv86sC88T0W31rd3c+EW6QZq6sRPGPhUcNIOhGf1fDKR9kSxqu3C+NK80OMT1BA==","signing_key_id":"pith-v1-2026-05"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-19T18:37:17Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3vlqyEs3Uk6WRxssAQETis6inUdLD2h9MI8KRld1H/RKjfzUPGJvdLM/Td0InG6KDPtMQo0jrVahHW27CTjpBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T03:19:51.952470Z"},"content_sha256":"25b3050d3e7229bdd8a1434453ed1f2ae42dcfcda90194b12fce92f18367d981","schema_version":"1.0","event_id":"sha256:25b3050d3e7229bdd8a1434453ed1f2ae42dcfcda90194b12fce92f18367d981"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/MEFZ47YPYK7CJ7SUKJUBDQ75NS/bundle.json","state_url":"https://pith.science/pith/MEFZ47YPYK7CJ7SUKJUBDQ75NS/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/MEFZ47YPYK7CJ7SUKJUBDQ75NS/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-05T03:19:51Z","links":{"resolver":"https://pith.science/pith/MEFZ47YPYK7CJ7SUKJUBDQ75NS","bundle":"https://pith.science/pith/MEFZ47YPYK7CJ7SUKJUBDQ75NS/bundle.json","state":"https://pith.science/pith/MEFZ47YPYK7CJ7SUKJUBDQ75NS/state.json","well_known_bundle":"https://pith.science/.well-known/pith/MEFZ47YPYK7CJ7SUKJUBDQ75NS/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:MEFZ47YPYK7CJ7SUKJUBDQ75NS","merge_version":"pith-open-graph-merge-v1","event_count":5,"valid_event_count":5,"invalid_event_count":0,"equivocation_count":1,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"821545c2ea291801ec636cb20017331dd9a91be57be4b261ceec495a923e3c1f","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-04-30T17:07:11Z","title_canon_sha256":"51cece6c498dc2d9af2b522f877e6dd9698e53e74233fe706d0fa66424f125b3"},"schema_version":"1.0","source":{"id":"2604.28118","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.28118","created_at":"2026-06-26T01:15:19Z"},{"alias_kind":"arxiv_version","alias_value":"2604.28118v2","created_at":"2026-06-26T01:15:19Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.28118","created_at":"2026-06-26T01:15:19Z"},{"alias_kind":"pith_short_12","alias_value":"MEFZ47YPYK7C","created_at":"2026-06-26T01:15:19Z"},{"alias_kind":"pith_short_16","alias_value":"MEFZ47YPYK7CJ7SU","created_at":"2026-06-26T01:15:19Z"},{"alias_kind":"pith_short_8","alias_value":"MEFZ47YP","created_at":"2026-06-26T01:15:19Z"}],"graph_snapshots":[{"event_id":"sha256:f4a2dc9537a305baadf7f7668c254be1e297ec8776eb07d7c7df1ef0ec046296","target":"graph","created_at":"2026-06-26T01:15:19Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"On DEFault-bench, DEFault++ exceeds an AUROC of 0.96 for detection and a Macro-F1 of 0.85 for both categorization and root-cause diagnosis on encoder and decoder architectures. In a developer study with 21 practitioners, the accuracy of choosing correct repair actions increased from 57.1% without support to 83.3% when using DEFault++."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The faults generated by the DEForm mutation technique accurately represent the kinds of faults that occur in real deployed transformer models, and the Fault Propagation Graph derived from the architecture fully captures how those faults affect observable behavior."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"DEFault++ delivers automated hierarchical fault detection, categorization into 12 transformer-specific types, and root-cause diagnosis among 45 mechanisms on a new benchmark of 3,739 mutated instances, with AUROC >0.96 and Macro-F1 0.85, plus improved developer repair accuracy in a user study."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"DEFault++ detects whether a fault exists in a transformer model, classifies it into one of 12 specific categories, and identifies its root cause among up to 45 mechanisms by analyzing runtime behaviors through an architecture-derived graph."}],"snapshot_sha256":"74b83fc509150acb52609245e25cf606d384188cca259d1f4fc71dfc4fc27bc2"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":false,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T20:40:38.128942Z","status":"completed","version":"1.0.0"},{"findings_count":3,"name":"doi_compliance","ran_at":"2026-05-19T18:36:39.919630Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.28118/integrity.json","findings":[{"audited_at":"2026-05-19T18:36:39.919630Z","detected_arxiv_id":null,"detected_doi":"10.1109/ase.2019.000","detector":"doi_compliance","finding_type":"unresolvable_identifier","note":"Identifier '10.1109/ase.2019.000' is syntactically valid but the DOI registry (doi.org) returned 404, and Crossref / OpenAlex / internal corpus also have no record. The cited work could not be located through any authoritative source.","ref_index":40,"severity":"critical","verdict_class":"cross_source"},{"audited_at":"2026-05-19T18:36:39.919630Z","detected_arxiv_id":null,"detected_doi":"10.1145/3180155.3180220.Ashish","detector":"doi_compliance","finding_type":"recoverable_identifier","note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1145/3180155.3180220.Ashish) was visible in the surrounding text but could not be confirmed against doi.org as printed.","ref_index":34,"severity":"advisory","verdict_class":"incontrovertible"},{"audited_at":"2026-05-19T18:36:39.919630Z","detected_arxiv_id":null,"detected_doi":"10.1109/TSE.2025.3543187.Ketai","detector":"doi_compliance","finding_type":"recoverable_identifier","note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1109/TSE.2025.3543187.Ketai) was visible in the surrounding text but could not be confirmed against doi.org as printed.","ref_index":27,"severity":"advisory","verdict_class":"incontrovertible"}],"snapshot_sha256":"9d7c299ad0f4190d94e5474087717ec45a38a25b17e28865ce4b20545c27be55","summary":{"advisory":2,"by_detector":{"doi_compliance":{"advisory":2,"critical":1,"informational":0,"total":3}},"critical":1,"informational":0}},"paper":{"abstract_excerpt":"Transformers now underpin critical AI systems across industry and research. Yet their faults can silently alter model behavior without runtime errors, and existing techniques offer little support for tracing these failures to their component and root cause. Such faults evade detection because loss and numerical values stay normal, and the visible symptom rarely identifies the component responsible. We present DEFault++, a hierarchical learning-based technique that first detects a fault, then identifies the affected component, and finally the cause within it, helping developers effectively debu","authors_text":"Mohammad Masudur Rahman, Saurabh Singh Rajput, Sigma Jahan, Tushar Sharma","cross_cats":["cs.AI","cs.LG"],"headline":"DEFault++ detects whether a fault exists in a transformer model, classifies it into one of 12 specific categories, and identifies its root cause among up to 45 mechanisms by analyzing runtime behaviors through an architecture-derived graph.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-04-30T17:07:11Z","title":"Hierarchical Fault Detection and Diagnosis for Transformer Architectures"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.28118","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-07T05:28:19.157068Z","id":"223c4a4d-aa2a-491d-8bb6-b22fac962e50","model_set":{"reader":"grok-4.3"},"one_line_summary":"DEFault++ delivers automated hierarchical fault detection, categorization into 12 transformer-specific types, and root-cause diagnosis among 45 mechanisms on a new benchmark of 3,739 mutated instances, with AUROC >0.96 and Macro-F1 0.85, plus improved developer repair accuracy in a user study.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"DEFault++ detects whether a fault exists in a transformer model, classifies it into one of 12 specific categories, and identifies its root cause among up to 45 mechanisms by analyzing runtime behaviors through an architecture-derived graph.","strongest_claim":"On DEFault-bench, DEFault++ exceeds an AUROC of 0.96 for detection and a Macro-F1 of 0.85 for both categorization and root-cause diagnosis on encoder and decoder architectures. In a developer study with 21 practitioners, the accuracy of choosing correct repair actions increased from 57.1% without support to 83.3% when using DEFault++.","weakest_assumption":"The faults generated by the DEForm mutation technique accurately represent the kinds of faults that occur in real deployed transformer models, and the Fault Propagation Graph derived from the architecture fully captures how those faults affect observable behavior."}},"verdict_id":"223c4a4d-aa2a-491d-8bb6-b22fac962e50"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:42ee59fbae50ccaf9ee10d31184ebfe32f1f32cc6046390baf546b19a9d57f3b","target":"record","created_at":"2026-06-26T01:15:19Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"821545c2ea291801ec636cb20017331dd9a91be57be4b261ceec495a923e3c1f","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-04-30T17:07:11Z","title_canon_sha256":"51cece6c498dc2d9af2b522f877e6dd9698e53e74233fe706d0fa66424f125b3"},"schema_version":"1.0","source":{"id":"2604.28118","kind":"arxiv","version":2}},"canonical_sha256":"610b9e7f0fc2be24fe54526811c3fd6cabb2efdb5e57da660681178afb9b51cd","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"610b9e7f0fc2be24fe54526811c3fd6cabb2efdb5e57da660681178afb9b51cd","first_computed_at":"2026-06-26T01:15:19.009534Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-26T01:15:19.009534Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"d9nSISqHVV3mZhmZCXlwEIpOeGhTQ7Q9iJovPm056BW569VL7r9HHOOsW/CfrRznLl0FjxSdjrVae+L1st7cAA==","signature_status":"signed_v1","signed_at":"2026-06-26T01:15:19.009964Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.28118","source_kind":"arxiv","source_version":2}}},"equivocations":[{"signer_id":"pith.science","event_type":"integrity_finding","target":"integrity","event_ids":["sha256:25b3050d3e7229bdd8a1434453ed1f2ae42dcfcda90194b12fce92f18367d981","sha256:4756a26d17ab3411eb7e5d1ada67e75428c0f930c07827e0e67fc26c96092fa3","sha256:f4c8bed3a4c4fdedd212eacdafbfc697a197320d3cfc576321f3da1c0b55a2d6"]}],"invalid_events":[],"applied_event_ids":["sha256:42ee59fbae50ccaf9ee10d31184ebfe32f1f32cc6046390baf546b19a9d57f3b","sha256:f4a2dc9537a305baadf7f7668c254be1e297ec8776eb07d7c7df1ef0ec046296"],"state_sha256":"b8b626cbded9fc96ae00cb490dd6be2f6d6cd1aae8388b3c9405608df6d444f3"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"zeuhjFj2/cYTBGirCL3SBy7EW+gr+rXQtBsRx/S+tMbHqV7f7fons1hmufaWyAOeHk93A6SzCKfV1G8YnmRdDg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T03:19:51.955033Z","bundle_sha256":"df063ff66784784566102282c7c5880300ecfb0a0cdfb20d2e5b62f8127f46e7"}}