{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7TSPPAKEPYJ2YWKZP752UU3ULO","short_pith_number":"pith:7TSPPAKE","schema_version":"1.0","canonical_sha256":"fce4f781447e13ac59597ffbaa53745bb1a8794a82c756db90ca4554e781fae5","source":{"kind":"arxiv","id":"2606.28116","version":1},"attestation_state":"computed","paper":{"title":"Mechanism-Driven Monitors for Preemptive Detection of LLM Training Instability","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ansheng You, Fan Wu, Hantao Huang, Ruixuan Huang, Shuai Wang, Wenyi Fang, Yang Zheng, Yifan Huang, Yipei Wang, Zhenxing Zhang","submitted_at":"2026-06-26T14:18:22Z","abstract_excerpt":"Frontier large language model training consumes massive accelerator fleets and long wall-clock computation, making stability failures costly when they occur. After a numerical or a hyperparameter fault has already destabilized the training dynamics, it may continue for thousands of steps while loss and gradient norms still appear normal. We study mechanism-driven detection of training instability by deriving internal monitors from the functional role of each critical module and from the earliest computational sites where failures are expected to produce measurable signatures. For low-precision"},"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":"2606.28116","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-26T14:18:22Z","cross_cats_sorted":[],"title_canon_sha256":"d6225740b9a3fb0cd8d0dcbfd65f1b819705389952185b07ee97e4279db7aa72","abstract_canon_sha256":"3a56021f95ed74b37727e25a1a7ba106a4c74ab45f79d41f411d73195258f303"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-29T01:15:06.508735Z","signature_b64":"S7cm1mUDS/YquL4auwO/6TNPgGk3TnwiDn04nxiYk8d15h4gcRL48GnXxSg27OQ6gx8Zzut9mGGi173r8VS6DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fce4f781447e13ac59597ffbaa53745bb1a8794a82c756db90ca4554e781fae5","last_reissued_at":"2026-06-29T01:15:06.508206Z","signature_status":"signed_v1","first_computed_at":"2026-06-29T01:15:06.508206Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mechanism-Driven Monitors for Preemptive Detection of LLM Training Instability","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ansheng You, Fan Wu, Hantao Huang, Ruixuan Huang, Shuai Wang, Wenyi Fang, Yang Zheng, Yifan Huang, Yipei Wang, Zhenxing Zhang","submitted_at":"2026-06-26T14:18:22Z","abstract_excerpt":"Frontier large language model training consumes massive accelerator fleets and long wall-clock computation, making stability failures costly when they occur. After a numerical or a hyperparameter fault has already destabilized the training dynamics, it may continue for thousands of steps while loss and gradient norms still appear normal. We study mechanism-driven detection of training instability by deriving internal monitors from the functional role of each critical module and from the earliest computational sites where failures are expected to produce measurable signatures. For low-precision"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.28116","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/2606.28116/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":"2606.28116","created_at":"2026-06-29T01:15:06.508256+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.28116v1","created_at":"2026-06-29T01:15:06.508256+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.28116","created_at":"2026-06-29T01:15:06.508256+00:00"},{"alias_kind":"pith_short_12","alias_value":"7TSPPAKEPYJ2","created_at":"2026-06-29T01:15:06.508256+00:00"},{"alias_kind":"pith_short_16","alias_value":"7TSPPAKEPYJ2YWKZ","created_at":"2026-06-29T01:15:06.508256+00:00"},{"alias_kind":"pith_short_8","alias_value":"7TSPPAKE","created_at":"2026-06-29T01:15:06.508256+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/7TSPPAKEPYJ2YWKZP752UU3ULO","json":"https://pith.science/pith/7TSPPAKEPYJ2YWKZP752UU3ULO.json","graph_json":"https://pith.science/api/pith-number/7TSPPAKEPYJ2YWKZP752UU3ULO/graph.json","events_json":"https://pith.science/api/pith-number/7TSPPAKEPYJ2YWKZP752UU3ULO/events.json","paper":"https://pith.science/paper/7TSPPAKE"},"agent_actions":{"view_html":"https://pith.science/pith/7TSPPAKEPYJ2YWKZP752UU3ULO","download_json":"https://pith.science/pith/7TSPPAKEPYJ2YWKZP752UU3ULO.json","view_paper":"https://pith.science/paper/7TSPPAKE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.28116&json=true","fetch_graph":"https://pith.science/api/pith-number/7TSPPAKEPYJ2YWKZP752UU3ULO/graph.json","fetch_events":"https://pith.science/api/pith-number/7TSPPAKEPYJ2YWKZP752UU3ULO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7TSPPAKEPYJ2YWKZP752UU3ULO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7TSPPAKEPYJ2YWKZP752UU3ULO/action/storage_attestation","attest_author":"https://pith.science/pith/7TSPPAKEPYJ2YWKZP752UU3ULO/action/author_attestation","sign_citation":"https://pith.science/pith/7TSPPAKEPYJ2YWKZP752UU3ULO/action/citation_signature","submit_replication":"https://pith.science/pith/7TSPPAKEPYJ2YWKZP752UU3ULO/action/replication_record"}},"created_at":"2026-06-29T01:15:06.508256+00:00","updated_at":"2026-06-29T01:15:06.508256+00:00"}