{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:UN2O3Z3ZNARDLNSIVDWXADDW6J","short_pith_number":"pith:UN2O3Z3Z","canonical_record":{"source":{"id":"2605.06915","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-07T20:25:02Z","cross_cats_sorted":[],"title_canon_sha256":"7fb1dafc4e1383768a754c89e1a14721f6c39c3d5acaf156f3369206f082bc50","abstract_canon_sha256":"9a640245a89cd0475aa1110ca504e1892737855942fd84819109d3f8d75cab0d"},"schema_version":"1.0"},"canonical_sha256":"a374ede779682235b648a8ed700c76f246378e96401b2db890275350035f5bfe","source":{"kind":"arxiv","id":"2605.06915","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.06915","created_at":"2026-05-28T01:04:41Z"},{"alias_kind":"arxiv_version","alias_value":"2605.06915v2","created_at":"2026-05-28T01:04:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.06915","created_at":"2026-05-28T01:04:41Z"},{"alias_kind":"pith_short_12","alias_value":"UN2O3Z3ZNARD","created_at":"2026-05-28T01:04:41Z"},{"alias_kind":"pith_short_16","alias_value":"UN2O3Z3ZNARDLNSI","created_at":"2026-05-28T01:04:41Z"},{"alias_kind":"pith_short_8","alias_value":"UN2O3Z3Z","created_at":"2026-05-28T01:04:41Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:UN2O3Z3ZNARDLNSIVDWXADDW6J","target":"record","payload":{"canonical_record":{"source":{"id":"2605.06915","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-07T20:25:02Z","cross_cats_sorted":[],"title_canon_sha256":"7fb1dafc4e1383768a754c89e1a14721f6c39c3d5acaf156f3369206f082bc50","abstract_canon_sha256":"9a640245a89cd0475aa1110ca504e1892737855942fd84819109d3f8d75cab0d"},"schema_version":"1.0"},"canonical_sha256":"a374ede779682235b648a8ed700c76f246378e96401b2db890275350035f5bfe","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:41.768957Z","signature_b64":"a4yu4ex3CgwCo2ukQ5DcJfNraPVbSYKxhEKBUPb7LbUzV4Hsx4AViaRBbgncoVw8yT4P3tD2q3gNtTfQ1llJCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a374ede779682235b648a8ed700c76f246378e96401b2db890275350035f5bfe","last_reissued_at":"2026-05-28T01:04:41.768137Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:41.768137Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.06915","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-28T01:04:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7MLoPxeJEl1bHsxNivz3/YIBH99le78v51ZwEqo0+P9TVnbhRL11Ww03NC/QxX0vTeDMJXWedtArVd7EdrT/BA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-30T07:07:20.035588Z"},"content_sha256":"03c5a00c9b38ef0b72539b036c8959393679a22338aa65c05dd46a8ec76738a6","schema_version":"1.0","event_id":"sha256:03c5a00c9b38ef0b72539b036c8959393679a22338aa65c05dd46a8ec76738a6"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:UN2O3Z3ZNARDLNSIVDWXADDW6J","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"LLMs are not (consistently) Bayesian: Quantifying internal (in)consistencies of LLMs' probabilistic beliefs","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"Large language models do not consistently update probabilistic beliefs according to Bayesian rules.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Adam Goli\\'nski, Chacha Chen, Guillermo Sapiro, Masha Fedzechkina, Matthew J\\\"orke, Nicholas Foti, Sinead Williamson","submitted_at":"2026-05-07T20:25:02Z","abstract_excerpt":"Modern AI systems are being deployed in complex domains such as medicine, science, and law, where it is important that they not only produce correct answers, but also represent and update uncertain beliefs about the world as new evidence arrives. We introduce the novel technique of studying LLMs as information processing rules and utilize the information processing gap to study the internal (in)consistencies of how LLMs update their probabilistic beliefs from evidence. Our extensive experiments evaluate multiple approaches in which LLMs can incorporate evidence into their beliefs. Some of thes"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Some of these approaches produce (nearly) Bayesian updates; others seem to use a learned heuristic. Surprisingly, the non-Bayesian heuristic updates often outperform exact Bayesian computation in terms of downstream task performance -- indicating the LLMs' probabilistic models of the world are misspecified.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That LLMs possess stable internal probabilistic beliefs that can be reliably elicited and compared to Bayesian standards via prompting or other interfaces, and that the introduced information processing gap accurately quantifies internal inconsistencies rather than surface-level response artifacts.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"LLMs do not consistently perform Bayesian updates on probabilistic beliefs; heuristic approaches often outperform exact Bayesian computation on downstream tasks, indicating misspecified internal models of the world.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Large language models do not consistently update probabilistic beliefs according to Bayesian rules.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"0f7904ca30032647b59b6cadb157f779c185474afebf4c45138b325e1c5482be"},"source":{"id":"2605.06915","kind":"arxiv","version":2},"verdict":{"id":"b93c9c90-c93e-4274-8f9c-abb0515f7aeb","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-11T00:49:03.024641Z","strongest_claim":"Some of these approaches produce (nearly) Bayesian updates; others seem to use a learned heuristic. Surprisingly, the non-Bayesian heuristic updates often outperform exact Bayesian computation in terms of downstream task performance -- indicating the LLMs' probabilistic models of the world are misspecified.","one_line_summary":"LLMs do not consistently perform Bayesian updates on probabilistic beliefs; heuristic approaches often outperform exact Bayesian computation on downstream tasks, indicating misspecified internal models of the world.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That LLMs possess stable internal probabilistic beliefs that can be reliably elicited and compared to Bayesian standards via prompting or other interfaces, and that the introduced information processing gap accurately quantifies internal inconsistencies rather than surface-level response artifacts.","pith_extraction_headline":"Large language models do not consistently update probabilistic beliefs according to Bayesian rules."},"integrity":{"clean":false,"summary":{"advisory":1,"critical":0,"by_detector":{"doi_compliance":{"total":1,"advisory":1,"critical":0,"informational":0}},"informational":0},"endpoint":"/pith/2605.06915/integrity.json","findings":[{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.18653/v1/2024.emnlp-main.153.URLhttps://aclanthology.org/2024.emnlp-main.153/.Omar) was visible in the surrounding text but could not be confirmed against doi.org as printed.","detector":"doi_compliance","severity":"advisory","ref_index":11,"audited_at":"2026-05-19T12:15:10.338357Z","detected_doi":"10.18653/v1/2024.emnlp-main.153.URLhttps://aclanthology.org/2024.emnlp-main.153/.Omar","finding_type":"recoverable_identifier","verdict_class":"incontrovertible","detected_arxiv_id":null}],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-20T11:42:03.535238Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-20T06:40:18.370483Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T17:31:19.390553Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T12:15:10.338357Z","status":"completed","version":"1.0.0","findings_count":1}],"snapshot_sha256":"73e4394f573496688d1cb75a181a1679dc0da129b30792ed07f05c6217c51493"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"4e7bf806b1f39a72dfc51e8e3c3a75037130c42c999cad517f3597cbf5523f4f"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"b93c9c90-c93e-4274-8f9c-abb0515f7aeb"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-28T01:04:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"rC2chHIosYZgfc6B+8sj6kkOAMK6/UgZJA6Jxi8fHk1aiFL8qS11klYLSBe5rUEx4FAuUIXyoYzlRujUp7BADg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-30T07:07:20.036155Z"},"content_sha256":"a905d46140a9cbf145bae84ab0a7a1bbaba6ac7d2427d72abd99eb37d33b6ec6","schema_version":"1.0","event_id":"sha256:a905d46140a9cbf145bae84ab0a7a1bbaba6ac7d2427d72abd99eb37d33b6ec6"},{"event_type":"integrity_finding","subject_pith_number":"pith:2026:UN2O3Z3ZNARDLNSIVDWXADDW6J","target":"integrity","payload":{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.18653/v1/2024.emnlp-main.153.URLhttps://aclanthology.org/2024.emnlp-main.153/.Omar) was visible in the surrounding text but could not be confirmed against doi.org as printed.","snippet":"Association for Computational Linguistics. doi: 10.18653/v1/ 2024.emnlp-main.153. URLhttps://aclanthology.org/2024.emnlp-main.153/. Omar Shaikh, Shardul Sapkota, Shan Rizvi, Eric Horvitz, Joon Sung Park, Diyi Yang, and Michael S Bernstein. ","arxiv_id":"2605.06915","detector":"doi_compliance","evidence":{"ref_index":11,"verdict_class":"incontrovertible","resolved_title":null,"printed_excerpt":"Association for Computational Linguistics. doi: 10.18653/v1/ 2024.emnlp-main.153. URLhttps://aclanthology.org/2024.emnlp-main.153/. Omar Shaikh, Shardul Sapkota, Shan Rizvi, Eric Horvitz, Joon Sung Park, Diyi Yang, and Michael S Bernstein. ","reconstructed_doi":"10.18653/v1/2024.emnlp-main.153.URLhttps://aclanthology.org/2024.emnlp-main.153/.Omar"},"severity":"advisory","ref_index":11,"audited_at":"2026-05-19T12:15:10.338357Z","event_type":"pith.integrity.v1","detected_doi":"10.18653/v1/2024.emnlp-main.153.URLhttps://aclanthology.org/2024.emnlp-main.153/.Omar","detector_url":"https://pith.science/pith-integrity-protocol#doi_compliance","external_url":null,"finding_type":"recoverable_identifier","evidence_hash":"15678eb1f320a8301c4dae8eb1a9a16f8a243e87ef399a5f6cfa40a42d99e3ca","paper_version":1,"verdict_class":"incontrovertible","resolved_title":null,"detector_version":"1.0.0","detected_arxiv_id":null,"integrity_event_id":1189,"payload_sha256":"a48b85468f95efeb52465b9bacfe95ed35a254c69cc43deace399b41eef661d0","signature_b64":"iba+M/ItbaieFclp2ne+Cy2YeSXuHHJ2d2BSVYvinix011cy+3NnEtPznkm1+1rIqp770/Rb78Mf6kGQzgzKDA==","signing_key_id":"pith-v1-2026-05"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-19T12:17:03Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"hPgaIUE9R8fI9jjTlPm494yrnpMc5f3xfxJcbEOtfh5cTUWw+xFE1wjQlFVEkisJJ/kBI+mhFMmvkUGOO4EAAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-30T07:07:20.037136Z"},"content_sha256":"1bf67f1ce24254192b2ee8bbd044f61df990ef3c7fd9166b34b584f51bd8615c","schema_version":"1.0","event_id":"sha256:1bf67f1ce24254192b2ee8bbd044f61df990ef3c7fd9166b34b584f51bd8615c"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/UN2O3Z3ZNARDLNSIVDWXADDW6J/bundle.json","state_url":"https://pith.science/pith/UN2O3Z3ZNARDLNSIVDWXADDW6J/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/UN2O3Z3ZNARDLNSIVDWXADDW6J/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-06-30T07:07:20Z","links":{"resolver":"https://pith.science/pith/UN2O3Z3ZNARDLNSIVDWXADDW6J","bundle":"https://pith.science/pith/UN2O3Z3ZNARDLNSIVDWXADDW6J/bundle.json","state":"https://pith.science/pith/UN2O3Z3ZNARDLNSIVDWXADDW6J/state.json","well_known_bundle":"https://pith.science/.well-known/pith/UN2O3Z3ZNARDLNSIVDWXADDW6J/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:UN2O3Z3ZNARDLNSIVDWXADDW6J","merge_version":"pith-open-graph-merge-v1","event_count":3,"valid_event_count":3,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"9a640245a89cd0475aa1110ca504e1892737855942fd84819109d3f8d75cab0d","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-07T20:25:02Z","title_canon_sha256":"7fb1dafc4e1383768a754c89e1a14721f6c39c3d5acaf156f3369206f082bc50"},"schema_version":"1.0","source":{"id":"2605.06915","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.06915","created_at":"2026-05-28T01:04:41Z"},{"alias_kind":"arxiv_version","alias_value":"2605.06915v2","created_at":"2026-05-28T01:04:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.06915","created_at":"2026-05-28T01:04:41Z"},{"alias_kind":"pith_short_12","alias_value":"UN2O3Z3ZNARD","created_at":"2026-05-28T01:04:41Z"},{"alias_kind":"pith_short_16","alias_value":"UN2O3Z3ZNARDLNSI","created_at":"2026-05-28T01:04:41Z"},{"alias_kind":"pith_short_8","alias_value":"UN2O3Z3Z","created_at":"2026-05-28T01:04:41Z"}],"graph_snapshots":[{"event_id":"sha256:a905d46140a9cbf145bae84ab0a7a1bbaba6ac7d2427d72abd99eb37d33b6ec6","target":"graph","created_at":"2026-05-28T01:04:41Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Some of these approaches produce (nearly) Bayesian updates; others seem to use a learned heuristic. Surprisingly, the non-Bayesian heuristic updates often outperform exact Bayesian computation in terms of downstream task performance -- indicating the LLMs' probabilistic models of the world are misspecified."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That LLMs possess stable internal probabilistic beliefs that can be reliably elicited and compared to Bayesian standards via prompting or other interfaces, and that the introduced information processing gap accurately quantifies internal inconsistencies rather than surface-level response artifacts."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"LLMs do not consistently perform Bayesian updates on probabilistic beliefs; heuristic approaches often outperform exact Bayesian computation on downstream tasks, indicating misspecified internal models of the world."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Large language models do not consistently update probabilistic beliefs according to Bayesian rules."}],"snapshot_sha256":"0f7904ca30032647b59b6cadb157f779c185474afebf4c45138b325e1c5482be"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"4e7bf806b1f39a72dfc51e8e3c3a75037130c42c999cad517f3597cbf5523f4f"},"integrity":{"available":true,"clean":false,"detectors_run":[{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-20T11:42:03.535238Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T06:40:18.370483Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T17:31:19.390553Z","status":"completed","version":"1.0.0"},{"findings_count":1,"name":"doi_compliance","ran_at":"2026-05-19T12:15:10.338357Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.06915/integrity.json","findings":[{"audited_at":"2026-05-19T12:15:10.338357Z","detected_arxiv_id":null,"detected_doi":"10.18653/v1/2024.emnlp-main.153.URLhttps://aclanthology.org/2024.emnlp-main.153/.Omar","detector":"doi_compliance","finding_type":"recoverable_identifier","note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.18653/v1/2024.emnlp-main.153.URLhttps://aclanthology.org/2024.emnlp-main.153/.Omar) was visible in the surrounding text but could not be confirmed against doi.org as printed.","ref_index":11,"severity":"advisory","verdict_class":"incontrovertible"}],"snapshot_sha256":"73e4394f573496688d1cb75a181a1679dc0da129b30792ed07f05c6217c51493","summary":{"advisory":1,"by_detector":{"doi_compliance":{"advisory":1,"critical":0,"informational":0,"total":1}},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Modern AI systems are being deployed in complex domains such as medicine, science, and law, where it is important that they not only produce correct answers, but also represent and update uncertain beliefs about the world as new evidence arrives. We introduce the novel technique of studying LLMs as information processing rules and utilize the information processing gap to study the internal (in)consistencies of how LLMs update their probabilistic beliefs from evidence. Our extensive experiments evaluate multiple approaches in which LLMs can incorporate evidence into their beliefs. Some of thes","authors_text":"Adam Goli\\'nski, Chacha Chen, Guillermo Sapiro, Masha Fedzechkina, Matthew J\\\"orke, Nicholas Foti, Sinead Williamson","cross_cats":[],"headline":"Large language models do not consistently update probabilistic beliefs according to Bayesian rules.","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-07T20:25:02Z","title":"LLMs are not (consistently) Bayesian: Quantifying internal (in)consistencies of LLMs' probabilistic beliefs"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.06915","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-11T00:49:03.024641Z","id":"b93c9c90-c93e-4274-8f9c-abb0515f7aeb","model_set":{"reader":"grok-4.3"},"one_line_summary":"LLMs do not consistently perform Bayesian updates on probabilistic beliefs; heuristic approaches often outperform exact Bayesian computation on downstream tasks, indicating misspecified internal models of the world.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Large language models do not consistently update probabilistic beliefs according to Bayesian rules.","strongest_claim":"Some of these approaches produce (nearly) Bayesian updates; others seem to use a learned heuristic. Surprisingly, the non-Bayesian heuristic updates often outperform exact Bayesian computation in terms of downstream task performance -- indicating the LLMs' probabilistic models of the world are misspecified.","weakest_assumption":"That LLMs possess stable internal probabilistic beliefs that can be reliably elicited and compared to Bayesian standards via prompting or other interfaces, and that the introduced information processing gap accurately quantifies internal inconsistencies rather than surface-level response artifacts."}},"verdict_id":"b93c9c90-c93e-4274-8f9c-abb0515f7aeb"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:03c5a00c9b38ef0b72539b036c8959393679a22338aa65c05dd46a8ec76738a6","target":"record","created_at":"2026-05-28T01:04:41Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"9a640245a89cd0475aa1110ca504e1892737855942fd84819109d3f8d75cab0d","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-07T20:25:02Z","title_canon_sha256":"7fb1dafc4e1383768a754c89e1a14721f6c39c3d5acaf156f3369206f082bc50"},"schema_version":"1.0","source":{"id":"2605.06915","kind":"arxiv","version":2}},"canonical_sha256":"a374ede779682235b648a8ed700c76f246378e96401b2db890275350035f5bfe","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a374ede779682235b648a8ed700c76f246378e96401b2db890275350035f5bfe","first_computed_at":"2026-05-28T01:04:41.768137Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-28T01:04:41.768137Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"a4yu4ex3CgwCo2ukQ5DcJfNraPVbSYKxhEKBUPb7LbUzV4Hsx4AViaRBbgncoVw8yT4P3tD2q3gNtTfQ1llJCg==","signature_status":"signed_v1","signed_at":"2026-05-28T01:04:41.768957Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.06915","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:1bf67f1ce24254192b2ee8bbd044f61df990ef3c7fd9166b34b584f51bd8615c","sha256:03c5a00c9b38ef0b72539b036c8959393679a22338aa65c05dd46a8ec76738a6","sha256:a905d46140a9cbf145bae84ab0a7a1bbaba6ac7d2427d72abd99eb37d33b6ec6"],"state_sha256":"cd6487bccf88a10b8c7597979f08fddca57278f5b134a202c55c9f396a16540c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"JbvCe9GwX9eVqT/M9szWXKO3zrgBY7cosnQbrvE3GxMqvqMr1l8OShToY6/QSmzgfYwTulRLoIFXWGnkCMvZAQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-30T07:07:20.039422Z","bundle_sha256":"7eb9cf0b48638af25a6a91141820601e91e0b600d574ddb2a597f21ffb029bf7"}}