{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:QK5XUEBMRYU7HM2LDYGBE7YXD5","short_pith_number":"pith:QK5XUEBM","schema_version":"1.0","canonical_sha256":"82bb7a102c8e29f3b34b1e0c127f171f7ebf5b3569fdee54208c0b95a3deea33","source":{"kind":"arxiv","id":"2508.14802","version":1},"attestation_state":"computed","paper":{"title":"Privileged Self-Access Matters for Introspection in AI","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Harvey Lederman, Jennifer Hu, Kyle Mahowald, Siyuan Song","submitted_at":"2025-08-20T15:52:34Z","abstract_excerpt":"Whether AI models can introspect is an increasingly important practical question. But there is no consensus on how introspection is to be defined. Beginning from a recently proposed ''lightweight'' definition, we argue instead for a thicker one. According to our proposal, introspection in AI is any process which yields information about internal states through a process more reliable than one with equal or lower computational cost available to a third party. Using experiments where LLMs reason about their internal temperature parameters, we show they can appear to have lightweight introspectio"},"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":"2508.14802","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-08-20T15:52:34Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"538aff98281b14dabf87e4f2f7e2966f86c8de95d6dc6dc3e57352337fbae5bc","abstract_canon_sha256":"e8e9c3497ff5badcf144efbbbb949cadfa6adcf15e5a8553b8289b37746ed553"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:56:44.671173Z","signature_b64":"tOvxVryOG4YGDedfaicuvIR09qo2GnbhI8U0tJgWhBxT6MSo87yDOzPtkkpNsIxjiOLYlNj0lmCPEQxdY5TiDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"82bb7a102c8e29f3b34b1e0c127f171f7ebf5b3569fdee54208c0b95a3deea33","last_reissued_at":"2026-07-05T11:56:44.670383Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:56:44.670383Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Privileged Self-Access Matters for Introspection in AI","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Harvey Lederman, Jennifer Hu, Kyle Mahowald, Siyuan Song","submitted_at":"2025-08-20T15:52:34Z","abstract_excerpt":"Whether AI models can introspect is an increasingly important practical question. But there is no consensus on how introspection is to be defined. Beginning from a recently proposed ''lightweight'' definition, we argue instead for a thicker one. According to our proposal, introspection in AI is any process which yields information about internal states through a process more reliable than one with equal or lower computational cost available to a third party. Using experiments where LLMs reason about their internal temperature parameters, we show they can appear to have lightweight introspectio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.14802","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/2508.14802/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":"2508.14802","created_at":"2026-07-05T11:56:44.670449+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.14802v1","created_at":"2026-07-05T11:56:44.670449+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.14802","created_at":"2026-07-05T11:56:44.670449+00:00"},{"alias_kind":"pith_short_12","alias_value":"QK5XUEBMRYU7","created_at":"2026-07-05T11:56:44.670449+00:00"},{"alias_kind":"pith_short_16","alias_value":"QK5XUEBMRYU7HM2L","created_at":"2026-07-05T11:56:44.670449+00:00"},{"alias_kind":"pith_short_8","alias_value":"QK5XUEBM","created_at":"2026-07-05T11:56:44.670449+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.26242","citing_title":"Can LLMs Introspect? A Reality Check","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16325","citing_title":"Phase Transitions in Driven Informational Systems: A Two-Field Perspective on Learning Theory and Non-Equilibrium Chemistry","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2509.13316","citing_title":"Do Activation Verbalization Methods Convey Privileged Information?","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01428","citing_title":"Hallucinations Undermine Trust; Metacognition is a Way Forward","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06422","citing_title":"When to Call an Apple Red: Humans Follow Introspective Rules, VLMs Don't","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QK5XUEBMRYU7HM2LDYGBE7YXD5","json":"https://pith.science/pith/QK5XUEBMRYU7HM2LDYGBE7YXD5.json","graph_json":"https://pith.science/api/pith-number/QK5XUEBMRYU7HM2LDYGBE7YXD5/graph.json","events_json":"https://pith.science/api/pith-number/QK5XUEBMRYU7HM2LDYGBE7YXD5/events.json","paper":"https://pith.science/paper/QK5XUEBM"},"agent_actions":{"view_html":"https://pith.science/pith/QK5XUEBMRYU7HM2LDYGBE7YXD5","download_json":"https://pith.science/pith/QK5XUEBMRYU7HM2LDYGBE7YXD5.json","view_paper":"https://pith.science/paper/QK5XUEBM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.14802&json=true","fetch_graph":"https://pith.science/api/pith-number/QK5XUEBMRYU7HM2LDYGBE7YXD5/graph.json","fetch_events":"https://pith.science/api/pith-number/QK5XUEBMRYU7HM2LDYGBE7YXD5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QK5XUEBMRYU7HM2LDYGBE7YXD5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QK5XUEBMRYU7HM2LDYGBE7YXD5/action/storage_attestation","attest_author":"https://pith.science/pith/QK5XUEBMRYU7HM2LDYGBE7YXD5/action/author_attestation","sign_citation":"https://pith.science/pith/QK5XUEBMRYU7HM2LDYGBE7YXD5/action/citation_signature","submit_replication":"https://pith.science/pith/QK5XUEBMRYU7HM2LDYGBE7YXD5/action/replication_record"}},"created_at":"2026-07-05T11:56:44.670449+00:00","updated_at":"2026-07-05T11:56:44.670449+00:00"}