{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:EQMMM6FJF22BF3HHNM6OVIMY4P","short_pith_number":"pith:EQMMM6FJ","canonical_record":{"source":{"id":"2604.20022","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-21T21:59:57Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"26ab8e4c96eb874a9fa9dab3844fec69a305ef9ee71c90093236a075c8e9e7bc","abstract_canon_sha256":"4f8db9369fe22c7321438f57c509168fffaee8b7e0c9e16a250a60ad34cb404e"},"schema_version":"1.0"},"canonical_sha256":"2418c678a92eb412ece76b3ceaa198e3dda189c58edf55ac74a9da69374ea331","source":{"kind":"arxiv","id":"2604.20022","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.20022","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"arxiv_version","alias_value":"2604.20022v2","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.20022","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"pith_short_12","alias_value":"EQMMM6FJF22B","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"pith_short_16","alias_value":"EQMMM6FJF22BF3HH","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"pith_short_8","alias_value":"EQMMM6FJ","created_at":"2026-05-20T02:05:43Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:EQMMM6FJF22BF3HHNM6OVIMY4P","target":"record","payload":{"canonical_record":{"source":{"id":"2604.20022","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-21T21:59:57Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"26ab8e4c96eb874a9fa9dab3844fec69a305ef9ee71c90093236a075c8e9e7bc","abstract_canon_sha256":"4f8db9369fe22c7321438f57c509168fffaee8b7e0c9e16a250a60ad34cb404e"},"schema_version":"1.0"},"canonical_sha256":"2418c678a92eb412ece76b3ceaa198e3dda189c58edf55ac74a9da69374ea331","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T02:05:43.437490Z","signature_b64":"w/9hfphz6xSMVxDc8hjAlaZU5CQIf5tJdkLguKpwW72y3geQDakyXZ/+0eSgm6B9KSRrgxzybHMVDpT60zaVBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2418c678a92eb412ece76b3ceaa198e3dda189c58edf55ac74a9da69374ea331","last_reissued_at":"2026-05-20T02:05:43.436516Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T02:05:43.436516Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.20022","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":"2026-05-20T02:05:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"PywgYEtJajKSBWRRBSKzz7mn6cwmzByuDrw/Do+DhuAie6FI5bpBQ/Xg/b3VZ+TS8GGfp7PvBrFk1Sai1GFVCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-25T05:52:54.675400Z"},"content_sha256":"cb39ce66d2765f7b14591804b36d182dfe8ccfdf38166493cf2341ac8192b926","schema_version":"1.0","event_id":"sha256:cb39ce66d2765f7b14591804b36d182dfe8ccfdf38166493cf2341ac8192b926"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:EQMMM6FJF22BF3HHNM6OVIMY4P","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"MoBayes: A Modular Bayesian Framework for Separating Reasoning from Language in Conversational Clinical Decision Support","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Separating language from reasoning in medical AI allows a Bayesian engine to deliver calibrated diagnosis and beat larger standalone models.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Akhil Arora, Alexandra Kulinkina, David Sasu, Fay Elhassan, Jiayi Ma, Julien Stalhandske, Lars Klein, Mary-Anne Hartley, Yena Chang, Yusuf Kesmen","submitted_at":"2026-04-21T21:59:57Z","abstract_excerpt":"Large language models (LLMs) are increasingly used for conversational clinical decision support, yet they conflate next token prediction with probabilistic decision making. We argue that this conflation reflects an architectural limitation: such systems lack explicit posterior tracking, controllable abstention thresholds, and auditable reasoning chains. We introduce MoBayes, a Modular Bayesian dialogue framework that separates reasoning from language. The LLM acts only as a language interface, parsing patient conversation into structured observations, while a Bayesian module performs probabili"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"This separation yields three properties no autonomous LLM can offer: calibrated selective diagnosis with a continuously adjustable accuracy-coverage tradeoff, a statistical separation gap where even a cheap sensor paired with the engine outperforms a frontier standalone model from the same family at a fraction of the cost, and robustness to adversarial patient communication styles that cause standalone doctors to collapse.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That an LLM used only as a sensor can reliably convert natural language into accurate structured evidence without systematic errors that the Bayesian engine cannot correct, and that the knowledge bases (empirical or LLM-generated) supply priors and likelihoods sufficient for the claimed performance gains.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"BMBE separates LLM language handling from a standalone Bayesian diagnostic engine, producing calibrated selective diagnosis, a performance gap over frontier LLMs, and robustness to adversarial inputs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Separating language from reasoning in medical AI allows a Bayesian engine to deliver calibrated diagnosis and beat larger standalone models.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"62b1d022d9cb73c01c6e2c23161b6be4a45351caf9b89302ad7ace09ac180818"},"source":{"id":"2604.20022","kind":"arxiv","version":2},"verdict":{"id":"4ac5ea1a-1213-4f86-bf8c-d9a6fcfb8027","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T02:24:33.074233Z","strongest_claim":"This separation yields three properties no autonomous LLM can offer: calibrated selective diagnosis with a continuously adjustable accuracy-coverage tradeoff, a statistical separation gap where even a cheap sensor paired with the engine outperforms a frontier standalone model from the same family at a fraction of the cost, and robustness to adversarial patient communication styles that cause standalone doctors to collapse.","one_line_summary":"BMBE separates LLM language handling from a standalone Bayesian diagnostic engine, producing calibrated selective diagnosis, a performance gap over frontier LLMs, and robustness to adversarial inputs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That an LLM used only as a sensor can reliably convert natural language into accurate structured evidence without systematic errors that the Bayesian engine cannot correct, and that the knowledge bases (empirical or LLM-generated) supply priors and likelihoods sufficient for the claimed performance gains.","pith_extraction_headline":"Separating language from reasoning in medical AI allows a Bayesian engine to deliver calibrated diagnosis and beat larger standalone models."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.20022/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"},"verdict_id":"4ac5ea1a-1213-4f86-bf8c-d9a6fcfb8027"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T02:05:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Ld2VXxSAhswg+XFvyU2zxHxb5L/N0Kbvl3hVdC4z1SCSzky7xhAks9KB/ghnQsjwG96q7zHKHS40PBadT6ucDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-25T05:52:54.676069Z"},"content_sha256":"689513575a331eb07f07383829156ba8bdbb13e180f36f52c30ae8e8a1792e00","schema_version":"1.0","event_id":"sha256:689513575a331eb07f07383829156ba8bdbb13e180f36f52c30ae8e8a1792e00"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/EQMMM6FJF22BF3HHNM6OVIMY4P/bundle.json","state_url":"https://pith.science/pith/EQMMM6FJF22BF3HHNM6OVIMY4P/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/EQMMM6FJF22BF3HHNM6OVIMY4P/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":"2026-05-25T05:52:54Z","links":{"resolver":"https://pith.science/pith/EQMMM6FJF22BF3HHNM6OVIMY4P","bundle":"https://pith.science/pith/EQMMM6FJF22BF3HHNM6OVIMY4P/bundle.json","state":"https://pith.science/pith/EQMMM6FJF22BF3HHNM6OVIMY4P/state.json","well_known_bundle":"https://pith.science/.well-known/pith/EQMMM6FJF22BF3HHNM6OVIMY4P/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:EQMMM6FJF22BF3HHNM6OVIMY4P","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"4f8db9369fe22c7321438f57c509168fffaee8b7e0c9e16a250a60ad34cb404e","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-21T21:59:57Z","title_canon_sha256":"26ab8e4c96eb874a9fa9dab3844fec69a305ef9ee71c90093236a075c8e9e7bc"},"schema_version":"1.0","source":{"id":"2604.20022","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.20022","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"arxiv_version","alias_value":"2604.20022v2","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.20022","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"pith_short_12","alias_value":"EQMMM6FJF22B","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"pith_short_16","alias_value":"EQMMM6FJF22BF3HH","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"pith_short_8","alias_value":"EQMMM6FJ","created_at":"2026-05-20T02:05:43Z"}],"graph_snapshots":[{"event_id":"sha256:689513575a331eb07f07383829156ba8bdbb13e180f36f52c30ae8e8a1792e00","target":"graph","created_at":"2026-05-20T02:05:43Z","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":"This separation yields three properties no autonomous LLM can offer: calibrated selective diagnosis with a continuously adjustable accuracy-coverage tradeoff, a statistical separation gap where even a cheap sensor paired with the engine outperforms a frontier standalone model from the same family at a fraction of the cost, and robustness to adversarial patient communication styles that cause standalone doctors to collapse."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That an LLM used only as a sensor can reliably convert natural language into accurate structured evidence without systematic errors that the Bayesian engine cannot correct, and that the knowledge bases (empirical or LLM-generated) supply priors and likelihoods sufficient for the claimed performance gains."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"BMBE separates LLM language handling from a standalone Bayesian diagnostic engine, producing calibrated selective diagnosis, a performance gap over frontier LLMs, and robustness to adversarial inputs."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Separating language from reasoning in medical AI allows a Bayesian engine to deliver calibrated diagnosis and beat larger standalone models."}],"snapshot_sha256":"62b1d022d9cb73c01c6e2c23161b6be4a45351caf9b89302ad7ace09ac180818"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.20022/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large language models (LLMs) are increasingly used for conversational clinical decision support, yet they conflate next token prediction with probabilistic decision making. We argue that this conflation reflects an architectural limitation: such systems lack explicit posterior tracking, controllable abstention thresholds, and auditable reasoning chains. We introduce MoBayes, a Modular Bayesian dialogue framework that separates reasoning from language. The LLM acts only as a language interface, parsing patient conversation into structured observations, while a Bayesian module performs probabili","authors_text":"Akhil Arora, Alexandra Kulinkina, David Sasu, Fay Elhassan, Jiayi Ma, Julien Stalhandske, Lars Klein, Mary-Anne Hartley, Yena Chang, Yusuf Kesmen","cross_cats":["cs.AI","cs.CL"],"headline":"Separating language from reasoning in medical AI allows a Bayesian engine to deliver calibrated diagnosis and beat larger standalone models.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-21T21:59:57Z","title":"MoBayes: A Modular Bayesian Framework for Separating Reasoning from Language in Conversational Clinical Decision Support"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.20022","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T02:24:33.074233Z","id":"4ac5ea1a-1213-4f86-bf8c-d9a6fcfb8027","model_set":{"reader":"grok-4.3"},"one_line_summary":"BMBE separates LLM language handling from a standalone Bayesian diagnostic engine, producing calibrated selective diagnosis, a performance gap over frontier LLMs, and robustness to adversarial inputs.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Separating language from reasoning in medical AI allows a Bayesian engine to deliver calibrated diagnosis and beat larger standalone models.","strongest_claim":"This separation yields three properties no autonomous LLM can offer: calibrated selective diagnosis with a continuously adjustable accuracy-coverage tradeoff, a statistical separation gap where even a cheap sensor paired with the engine outperforms a frontier standalone model from the same family at a fraction of the cost, and robustness to adversarial patient communication styles that cause standalone doctors to collapse.","weakest_assumption":"That an LLM used only as a sensor can reliably convert natural language into accurate structured evidence without systematic errors that the Bayesian engine cannot correct, and that the knowledge bases (empirical or LLM-generated) supply priors and likelihoods sufficient for the claimed performance gains."}},"verdict_id":"4ac5ea1a-1213-4f86-bf8c-d9a6fcfb8027"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:cb39ce66d2765f7b14591804b36d182dfe8ccfdf38166493cf2341ac8192b926","target":"record","created_at":"2026-05-20T02:05:43Z","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":"4f8db9369fe22c7321438f57c509168fffaee8b7e0c9e16a250a60ad34cb404e","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-21T21:59:57Z","title_canon_sha256":"26ab8e4c96eb874a9fa9dab3844fec69a305ef9ee71c90093236a075c8e9e7bc"},"schema_version":"1.0","source":{"id":"2604.20022","kind":"arxiv","version":2}},"canonical_sha256":"2418c678a92eb412ece76b3ceaa198e3dda189c58edf55ac74a9da69374ea331","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2418c678a92eb412ece76b3ceaa198e3dda189c58edf55ac74a9da69374ea331","first_computed_at":"2026-05-20T02:05:43.436516Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T02:05:43.436516Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"w/9hfphz6xSMVxDc8hjAlaZU5CQIf5tJdkLguKpwW72y3geQDakyXZ/+0eSgm6B9KSRrgxzybHMVDpT60zaVBg==","signature_status":"signed_v1","signed_at":"2026-05-20T02:05:43.437490Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.20022","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:cb39ce66d2765f7b14591804b36d182dfe8ccfdf38166493cf2341ac8192b926","sha256:689513575a331eb07f07383829156ba8bdbb13e180f36f52c30ae8e8a1792e00"],"state_sha256":"936e5f6e19bafa4f4d8ba4abd44f2460299c1d43c7d219f32481c942e7cc412e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AnoLtN70UKSJXVDH/KljlPFOX21g0XZiJ7gjLAYzKVe86LsB/FmojiG8jqP7SblBuJSn396QToz9hnTPjmNSBQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-25T05:52:54.678876Z","bundle_sha256":"098a0a4e128cae867842eae3a1958cef31f3663495dd3377d03297161f89d592"}}