{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:T2JDCCJLYHLASWIVK6IUR4FZJH","short_pith_number":"pith:T2JDCCJL","schema_version":"1.0","canonical_sha256":"9e9231092bc1d6095915579148f0b949ea28b79739d7f826043feec4a0264bda","source":{"kind":"arxiv","id":"2606.25524","version":1},"attestation_state":"computed","paper":{"title":"Cliff Tokens: Identifying Single-Token Failure Triggers in LLM Mathematical Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Jaeyong Ko, Pilsung Kang, Yukyung Lee","submitted_at":"2026-06-24T08:03:24Z","abstract_excerpt":"Large language models (LLMs) reach high accuracy in mathematical reasoning, but individual traces on the same problem diverge; some arrive at the correct answer while others fail. Prior work analyzes failure at the step, chunk, or sentence level, or at tokens where failure has already occurred. Neither identifies the precise token that triggers the shift toward failure. We introduce the cliff token, a token where the token-wise potential drops significantly under an adaptive threshold that scales with the local token-wise potential, based on a one-sided two-proportion z-test. Across seven mode"},"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":"2606.25524","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-06-24T08:03:24Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"7188e536520d886a10b65f5c44dbf3544ae06cbe47b2069d148a6d8eea986300","abstract_canon_sha256":"311dab5cae06d881329a24a378a2489dc0f6be926d6f269f51909119a97e55d6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-25T01:18:07.828494Z","signature_b64":"ZPIOpqONq0DmSavGfTcUNFQJkE5uywIONQe3kFqrHB/PtqWbJpMyRgbPyt7F1u5BjYRyPdZ2uFZPPOWLOG5DAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9e9231092bc1d6095915579148f0b949ea28b79739d7f826043feec4a0264bda","last_reissued_at":"2026-06-25T01:18:07.828062Z","signature_status":"signed_v1","first_computed_at":"2026-06-25T01:18:07.828062Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cliff Tokens: Identifying Single-Token Failure Triggers in LLM Mathematical Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Jaeyong Ko, Pilsung Kang, Yukyung Lee","submitted_at":"2026-06-24T08:03:24Z","abstract_excerpt":"Large language models (LLMs) reach high accuracy in mathematical reasoning, but individual traces on the same problem diverge; some arrive at the correct answer while others fail. Prior work analyzes failure at the step, chunk, or sentence level, or at tokens where failure has already occurred. Neither identifies the precise token that triggers the shift toward failure. We introduce the cliff token, a token where the token-wise potential drops significantly under an adaptive threshold that scales with the local token-wise potential, based on a one-sided two-proportion z-test. Across seven mode"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.25524","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/2606.25524/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":"2606.25524","created_at":"2026-06-25T01:18:07.828121+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.25524v1","created_at":"2026-06-25T01:18:07.828121+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.25524","created_at":"2026-06-25T01:18:07.828121+00:00"},{"alias_kind":"pith_short_12","alias_value":"T2JDCCJLYHLA","created_at":"2026-06-25T01:18:07.828121+00:00"},{"alias_kind":"pith_short_16","alias_value":"T2JDCCJLYHLASWIV","created_at":"2026-06-25T01:18:07.828121+00:00"},{"alias_kind":"pith_short_8","alias_value":"T2JDCCJL","created_at":"2026-06-25T01:18:07.828121+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/T2JDCCJLYHLASWIVK6IUR4FZJH","json":"https://pith.science/pith/T2JDCCJLYHLASWIVK6IUR4FZJH.json","graph_json":"https://pith.science/api/pith-number/T2JDCCJLYHLASWIVK6IUR4FZJH/graph.json","events_json":"https://pith.science/api/pith-number/T2JDCCJLYHLASWIVK6IUR4FZJH/events.json","paper":"https://pith.science/paper/T2JDCCJL"},"agent_actions":{"view_html":"https://pith.science/pith/T2JDCCJLYHLASWIVK6IUR4FZJH","download_json":"https://pith.science/pith/T2JDCCJLYHLASWIVK6IUR4FZJH.json","view_paper":"https://pith.science/paper/T2JDCCJL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.25524&json=true","fetch_graph":"https://pith.science/api/pith-number/T2JDCCJLYHLASWIVK6IUR4FZJH/graph.json","fetch_events":"https://pith.science/api/pith-number/T2JDCCJLYHLASWIVK6IUR4FZJH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T2JDCCJLYHLASWIVK6IUR4FZJH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T2JDCCJLYHLASWIVK6IUR4FZJH/action/storage_attestation","attest_author":"https://pith.science/pith/T2JDCCJLYHLASWIVK6IUR4FZJH/action/author_attestation","sign_citation":"https://pith.science/pith/T2JDCCJLYHLASWIVK6IUR4FZJH/action/citation_signature","submit_replication":"https://pith.science/pith/T2JDCCJLYHLASWIVK6IUR4FZJH/action/replication_record"}},"created_at":"2026-06-25T01:18:07.828121+00:00","updated_at":"2026-06-25T01:18:07.828121+00:00"}