{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EVLMNV2SN22WXTTMPZHSYRCJLR","short_pith_number":"pith:EVLMNV2S","schema_version":"1.0","canonical_sha256":"2556c6d7526eb56bce6c7e4f2c44495c5d82baebb9b72a9e6cc42302c98899ed","source":{"kind":"arxiv","id":"2605.25267","version":1},"attestation_state":"computed","paper":{"title":"Latent Q-Barrier Shielding for Safe In-Context Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Amir Moeini, Lu Feng, Minjae Kwon, Shangtong Zhang","submitted_at":"2026-05-24T21:45:28Z","abstract_excerpt":"Safe in-context reinforcement learning (ICRL) adapts online from interaction history without test-time parameter updates while controlling episode cost under a safety budget. Under out-of-distribution (OOD) deployment shifts, pretraining-only safe ICRL can give poor reward-safety tradeoffs because the remaining budget affects behavior only through frozen policy conditioning, not an explicit action-level check against predicted future cost. We propose a latent Q-Barrier shield that learns a context representation, latent dynamics, and an ensemble cost critic before deployment. Without parameter"},"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":"2605.25267","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-24T21:45:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"9f1fd48fa173fc1888a3676c0874ace5e1205a1a15add982800f532b460c3877","abstract_canon_sha256":"d18ac02df8a44e921d451fac570d9e4c5261c3af468ad372b1ec010a77396e89"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T02:04:26.336536Z","signature_b64":"1fvfxHIQocU9MChCRvbK71+uNnPr8+SXdpznoMiReWusFXxTon2BTf3ngq+OgQM7sx+y+zZaLvbkhT/SjTZ4Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2556c6d7526eb56bce6c7e4f2c44495c5d82baebb9b72a9e6cc42302c98899ed","last_reissued_at":"2026-05-26T02:04:26.335715Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T02:04:26.335715Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Latent Q-Barrier Shielding for Safe In-Context Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Amir Moeini, Lu Feng, Minjae Kwon, Shangtong Zhang","submitted_at":"2026-05-24T21:45:28Z","abstract_excerpt":"Safe in-context reinforcement learning (ICRL) adapts online from interaction history without test-time parameter updates while controlling episode cost under a safety budget. Under out-of-distribution (OOD) deployment shifts, pretraining-only safe ICRL can give poor reward-safety tradeoffs because the remaining budget affects behavior only through frozen policy conditioning, not an explicit action-level check against predicted future cost. We propose a latent Q-Barrier shield that learns a context representation, latent dynamics, and an ensemble cost critic before deployment. Without parameter"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.25267","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/2605.25267/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":"2605.25267","created_at":"2026-05-26T02:04:26.335851+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.25267v1","created_at":"2026-05-26T02:04:26.335851+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.25267","created_at":"2026-05-26T02:04:26.335851+00:00"},{"alias_kind":"pith_short_12","alias_value":"EVLMNV2SN22W","created_at":"2026-05-26T02:04:26.335851+00:00"},{"alias_kind":"pith_short_16","alias_value":"EVLMNV2SN22WXTTM","created_at":"2026-05-26T02:04:26.335851+00:00"},{"alias_kind":"pith_short_8","alias_value":"EVLMNV2S","created_at":"2026-05-26T02:04:26.335851+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/EVLMNV2SN22WXTTMPZHSYRCJLR","json":"https://pith.science/pith/EVLMNV2SN22WXTTMPZHSYRCJLR.json","graph_json":"https://pith.science/api/pith-number/EVLMNV2SN22WXTTMPZHSYRCJLR/graph.json","events_json":"https://pith.science/api/pith-number/EVLMNV2SN22WXTTMPZHSYRCJLR/events.json","paper":"https://pith.science/paper/EVLMNV2S"},"agent_actions":{"view_html":"https://pith.science/pith/EVLMNV2SN22WXTTMPZHSYRCJLR","download_json":"https://pith.science/pith/EVLMNV2SN22WXTTMPZHSYRCJLR.json","view_paper":"https://pith.science/paper/EVLMNV2S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.25267&json=true","fetch_graph":"https://pith.science/api/pith-number/EVLMNV2SN22WXTTMPZHSYRCJLR/graph.json","fetch_events":"https://pith.science/api/pith-number/EVLMNV2SN22WXTTMPZHSYRCJLR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EVLMNV2SN22WXTTMPZHSYRCJLR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EVLMNV2SN22WXTTMPZHSYRCJLR/action/storage_attestation","attest_author":"https://pith.science/pith/EVLMNV2SN22WXTTMPZHSYRCJLR/action/author_attestation","sign_citation":"https://pith.science/pith/EVLMNV2SN22WXTTMPZHSYRCJLR/action/citation_signature","submit_replication":"https://pith.science/pith/EVLMNV2SN22WXTTMPZHSYRCJLR/action/replication_record"}},"created_at":"2026-05-26T02:04:26.335851+00:00","updated_at":"2026-05-26T02:04:26.335851+00:00"}