{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:X223NSX42FHTNXJULXZPBYQKKP","short_pith_number":"pith:X223NSX4","schema_version":"1.0","canonical_sha256":"beb5b6cafcd14f36dd345df2f0e20a53e3b6db521d8bdc7e262b07ead56da22d","source":{"kind":"arxiv","id":"2506.18781","version":1},"attestation_state":"computed","paper":{"title":"Existing LLMs Are Not Self-Consistent For Simple Tasks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Andrew Chi-Chih Yao, Jiawen Tao, Yang Yuan, Zhenru Lin","submitted_at":"2025-06-23T15:50:21Z","abstract_excerpt":"Large Language Models (LLMs) have grown increasingly powerful, yet ensuring their decisions remain transparent and trustworthy requires self-consistency -- no contradictions in their internal reasoning. Our study reveals that even on simple tasks, such as comparing points on a line or a plane, or reasoning in a family tree, all smaller models are highly inconsistent, and even state-of-the-art models like DeepSeek-R1 and GPT-o4-mini are not fully self-consistent. To quantify and mitigate these inconsistencies, we introduce inconsistency metrics and propose two automated methods -- a graph-based"},"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":"2506.18781","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-23T15:50:21Z","cross_cats_sorted":[],"title_canon_sha256":"d41553fea725b7b6aed6a0a6396dbd6a252e1cba2cc33bd5a21ab0b2b2a6f7eb","abstract_canon_sha256":"e20a4a5c4c9011698e93716274709c14cf86e2b2abbd758c885330cdaab4db36"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:25:55.976499Z","signature_b64":"guWwm3FJCLST32u6VJwU9zACsXYAUDUpPH9HDdKXl1bUnBn/6fyc2wWtOe1Ayq/iX+DU5jJKJJCJy2WZhsnXDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"beb5b6cafcd14f36dd345df2f0e20a53e3b6db521d8bdc7e262b07ead56da22d","last_reissued_at":"2026-07-05T11:25:55.976020Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:25:55.976020Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Existing LLMs Are Not Self-Consistent For Simple Tasks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Andrew Chi-Chih Yao, Jiawen Tao, Yang Yuan, Zhenru Lin","submitted_at":"2025-06-23T15:50:21Z","abstract_excerpt":"Large Language Models (LLMs) have grown increasingly powerful, yet ensuring their decisions remain transparent and trustworthy requires self-consistency -- no contradictions in their internal reasoning. Our study reveals that even on simple tasks, such as comparing points on a line or a plane, or reasoning in a family tree, all smaller models are highly inconsistent, and even state-of-the-art models like DeepSeek-R1 and GPT-o4-mini are not fully self-consistent. To quantify and mitigate these inconsistencies, we introduce inconsistency metrics and propose two automated methods -- a graph-based"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.18781","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/2506.18781/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":"2506.18781","created_at":"2026-07-05T11:25:55.976076+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.18781v1","created_at":"2026-07-05T11:25:55.976076+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.18781","created_at":"2026-07-05T11:25:55.976076+00:00"},{"alias_kind":"pith_short_12","alias_value":"X223NSX42FHT","created_at":"2026-07-05T11:25:55.976076+00:00"},{"alias_kind":"pith_short_16","alias_value":"X223NSX42FHTNXJU","created_at":"2026-07-05T11:25:55.976076+00:00"},{"alias_kind":"pith_short_8","alias_value":"X223NSX4","created_at":"2026-07-05T11:25:55.976076+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.05188","citing_title":"Position: It's Time to Optimize LLMs for Self-Consistency","ref_index":41,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/X223NSX42FHTNXJULXZPBYQKKP","json":"https://pith.science/pith/X223NSX42FHTNXJULXZPBYQKKP.json","graph_json":"https://pith.science/api/pith-number/X223NSX42FHTNXJULXZPBYQKKP/graph.json","events_json":"https://pith.science/api/pith-number/X223NSX42FHTNXJULXZPBYQKKP/events.json","paper":"https://pith.science/paper/X223NSX4"},"agent_actions":{"view_html":"https://pith.science/pith/X223NSX42FHTNXJULXZPBYQKKP","download_json":"https://pith.science/pith/X223NSX42FHTNXJULXZPBYQKKP.json","view_paper":"https://pith.science/paper/X223NSX4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.18781&json=true","fetch_graph":"https://pith.science/api/pith-number/X223NSX42FHTNXJULXZPBYQKKP/graph.json","fetch_events":"https://pith.science/api/pith-number/X223NSX42FHTNXJULXZPBYQKKP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/X223NSX42FHTNXJULXZPBYQKKP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/X223NSX42FHTNXJULXZPBYQKKP/action/storage_attestation","attest_author":"https://pith.science/pith/X223NSX42FHTNXJULXZPBYQKKP/action/author_attestation","sign_citation":"https://pith.science/pith/X223NSX42FHTNXJULXZPBYQKKP/action/citation_signature","submit_replication":"https://pith.science/pith/X223NSX42FHTNXJULXZPBYQKKP/action/replication_record"}},"created_at":"2026-07-05T11:25:55.976076+00:00","updated_at":"2026-07-05T11:25:55.976076+00:00"}