{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:O7GMQNXT4HAPP73SJERAA2KRVJ","short_pith_number":"pith:O7GMQNXT","schema_version":"1.0","canonical_sha256":"77ccc836f3e1c0f7ff724922006951aa6627e0ad79c1b00bb44d9e4e2d47d154","source":{"kind":"arxiv","id":"2607.18804","version":1},"attestation_state":"computed","paper":{"title":"Elicitation without Backpropagation: Steering Model Behavior by Optimizing the Latent Posterior","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Daniel Murfet, Garrett Baker, Susan Wei, Vinayak Pathak","submitted_at":"2026-07-21T07:34:18Z","abstract_excerpt":"In the \\emph{latent posterior model} of transformer behavior, the next-token distribution arises from a posterior over latent predictive models conditioned on the context, mixed to generate continuations. We exploit this model in settings where it is exact, namely Bayes-filtered transformers (BFTs) meta-learned on sequences from a hierarchical prior, to introduce \\textbf{Posterior Prefix Tuning (PPT)}, a new method for \\emph{eliciting} behavior from a transformer: given a utility function on continuations, find a prompt under which the transformer generates continuations of high expected utili"},"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":"2607.18804","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-21T07:34:18Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"793606561134d5da84d83fec9b4d1fd23adf7064453f2947af4e7ef6933055b6","abstract_canon_sha256":"8324f1c1d19f9413b227a9c1fd6c82445ca8fea44e9cd2bb5d65213348deea33"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-22T01:23:13.393242Z","signature_b64":"4r8NT4pjV3themo5Tw6LWCVPOCjjdSIALZX1dmgxAs++Jg2XrKXDeA7lvpQdTup5Hf0i5IU+hZAjkIQQtlDnCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"77ccc836f3e1c0f7ff724922006951aa6627e0ad79c1b00bb44d9e4e2d47d154","last_reissued_at":"2026-07-22T01:23:13.392396Z","signature_status":"signed_v1","first_computed_at":"2026-07-22T01:23:13.392396Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Elicitation without Backpropagation: Steering Model Behavior by Optimizing the Latent Posterior","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Daniel Murfet, Garrett Baker, Susan Wei, Vinayak Pathak","submitted_at":"2026-07-21T07:34:18Z","abstract_excerpt":"In the \\emph{latent posterior model} of transformer behavior, the next-token distribution arises from a posterior over latent predictive models conditioned on the context, mixed to generate continuations. We exploit this model in settings where it is exact, namely Bayes-filtered transformers (BFTs) meta-learned on sequences from a hierarchical prior, to introduce \\textbf{Posterior Prefix Tuning (PPT)}, a new method for \\emph{eliciting} behavior from a transformer: given a utility function on continuations, find a prompt under which the transformer generates continuations of high expected utili"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.18804","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/2607.18804/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":"2607.18804","created_at":"2026-07-22T01:23:13.392843+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.18804v1","created_at":"2026-07-22T01:23:13.392843+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.18804","created_at":"2026-07-22T01:23:13.392843+00:00"},{"alias_kind":"pith_short_12","alias_value":"O7GMQNXT4HAP","created_at":"2026-07-22T01:23:13.392843+00:00"},{"alias_kind":"pith_short_16","alias_value":"O7GMQNXT4HAPP73S","created_at":"2026-07-22T01:23:13.392843+00:00"},{"alias_kind":"pith_short_8","alias_value":"O7GMQNXT","created_at":"2026-07-22T01:23:13.392843+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/O7GMQNXT4HAPP73SJERAA2KRVJ","json":"https://pith.science/pith/O7GMQNXT4HAPP73SJERAA2KRVJ.json","graph_json":"https://pith.science/api/pith-number/O7GMQNXT4HAPP73SJERAA2KRVJ/graph.json","events_json":"https://pith.science/api/pith-number/O7GMQNXT4HAPP73SJERAA2KRVJ/events.json","paper":"https://pith.science/paper/O7GMQNXT"},"agent_actions":{"view_html":"https://pith.science/pith/O7GMQNXT4HAPP73SJERAA2KRVJ","download_json":"https://pith.science/pith/O7GMQNXT4HAPP73SJERAA2KRVJ.json","view_paper":"https://pith.science/paper/O7GMQNXT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.18804&json=true","fetch_graph":"https://pith.science/api/pith-number/O7GMQNXT4HAPP73SJERAA2KRVJ/graph.json","fetch_events":"https://pith.science/api/pith-number/O7GMQNXT4HAPP73SJERAA2KRVJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/O7GMQNXT4HAPP73SJERAA2KRVJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/O7GMQNXT4HAPP73SJERAA2KRVJ/action/storage_attestation","attest_author":"https://pith.science/pith/O7GMQNXT4HAPP73SJERAA2KRVJ/action/author_attestation","sign_citation":"https://pith.science/pith/O7GMQNXT4HAPP73SJERAA2KRVJ/action/citation_signature","submit_replication":"https://pith.science/pith/O7GMQNXT4HAPP73SJERAA2KRVJ/action/replication_record"}},"created_at":"2026-07-22T01:23:13.392843+00:00","updated_at":"2026-07-22T01:23:13.392843+00:00"}