{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:6QFJBFQCM6FLLMMWCEBGMTXYLV","short_pith_number":"pith:6QFJBFQC","schema_version":"1.0","canonical_sha256":"f40a909602678ab5b1961102664ef85d42de30b375cc8808895eb3c825b27fdc","source":{"kind":"arxiv","id":"2401.07927","version":4},"attestation_state":"computed","paper":{"title":"Are self-explanations from Large Language Models faithful?","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Andreas Madsen, Sarath Chandar, Siva Reddy","submitted_at":"2024-01-15T19:39:15Z","abstract_excerpt":"Instruction-tuned Large Language Models (LLMs) excel at many tasks and will even explain their reasoning, so-called self-explanations. However, convincing and wrong self-explanations can lead to unsupported confidence in LLMs, thus increasing risk. Therefore, it's important to measure if self-explanations truly reflect the model's behavior. Such a measure is called interpretability-faithfulness and is challenging to perform since the ground truth is inaccessible, and many LLMs only have an inference API. To address this, we propose employing self-consistency checks to measure faithfulness. For"},"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":"2401.07927","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2024-01-15T19:39:15Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"8509a05a7573a2728b562bc77b082ecd5574afe6e5c9d2379736859f0292d9a3","abstract_canon_sha256":"40aade2d6eb9b10cb68c05171f39b46fdb2b1096c2a95776953e6876105f280c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:19:57.939153Z","signature_b64":"fdXueT98YPTVBhYK4QyKoRLKkB3SS/aIN6BtyytxdG/OmNWO/EWasq12T1n0jP8DZFZzot0HeDeuE/2qQqE+DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f40a909602678ab5b1961102664ef85d42de30b375cc8808895eb3c825b27fdc","last_reissued_at":"2026-07-05T08:19:57.938650Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:19:57.938650Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Are self-explanations from Large Language Models faithful?","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Andreas Madsen, Sarath Chandar, Siva Reddy","submitted_at":"2024-01-15T19:39:15Z","abstract_excerpt":"Instruction-tuned Large Language Models (LLMs) excel at many tasks and will even explain their reasoning, so-called self-explanations. However, convincing and wrong self-explanations can lead to unsupported confidence in LLMs, thus increasing risk. Therefore, it's important to measure if self-explanations truly reflect the model's behavior. Such a measure is called interpretability-faithfulness and is challenging to perform since the ground truth is inaccessible, and many LLMs only have an inference API. To address this, we propose employing self-consistency checks to measure faithfulness. For"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.07927","kind":"arxiv","version":4},"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/2401.07927/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":"2401.07927","created_at":"2026-07-05T08:19:57.938719+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.07927v4","created_at":"2026-07-05T08:19:57.938719+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.07927","created_at":"2026-07-05T08:19:57.938719+00:00"},{"alias_kind":"pith_short_12","alias_value":"6QFJBFQCM6FL","created_at":"2026-07-05T08:19:57.938719+00:00"},{"alias_kind":"pith_short_16","alias_value":"6QFJBFQCM6FLLMMW","created_at":"2026-07-05T08:19:57.938719+00:00"},{"alias_kind":"pith_short_8","alias_value":"6QFJBFQC","created_at":"2026-07-05T08:19:57.938719+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2410.03296","citing_title":"A Systematic Comparison between Extractive Self-Explanations and Human Rationales in Text Classification","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2602.04003","citing_title":"When AI Persuades: Adversarial Explanation Attacks on Human Trust in AI-Assisted Decision Making","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14641","citing_title":"Learning to Draw ASCII Improves Spatial Reasoning in Language Models","ref_index":3,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6QFJBFQCM6FLLMMWCEBGMTXYLV","json":"https://pith.science/pith/6QFJBFQCM6FLLMMWCEBGMTXYLV.json","graph_json":"https://pith.science/api/pith-number/6QFJBFQCM6FLLMMWCEBGMTXYLV/graph.json","events_json":"https://pith.science/api/pith-number/6QFJBFQCM6FLLMMWCEBGMTXYLV/events.json","paper":"https://pith.science/paper/6QFJBFQC"},"agent_actions":{"view_html":"https://pith.science/pith/6QFJBFQCM6FLLMMWCEBGMTXYLV","download_json":"https://pith.science/pith/6QFJBFQCM6FLLMMWCEBGMTXYLV.json","view_paper":"https://pith.science/paper/6QFJBFQC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.07927&json=true","fetch_graph":"https://pith.science/api/pith-number/6QFJBFQCM6FLLMMWCEBGMTXYLV/graph.json","fetch_events":"https://pith.science/api/pith-number/6QFJBFQCM6FLLMMWCEBGMTXYLV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6QFJBFQCM6FLLMMWCEBGMTXYLV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6QFJBFQCM6FLLMMWCEBGMTXYLV/action/storage_attestation","attest_author":"https://pith.science/pith/6QFJBFQCM6FLLMMWCEBGMTXYLV/action/author_attestation","sign_citation":"https://pith.science/pith/6QFJBFQCM6FLLMMWCEBGMTXYLV/action/citation_signature","submit_replication":"https://pith.science/pith/6QFJBFQCM6FLLMMWCEBGMTXYLV/action/replication_record"}},"created_at":"2026-07-05T08:19:57.938719+00:00","updated_at":"2026-07-05T08:19:57.938719+00:00"}