{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:MS46TU7AT7NWFJXSCCLLT55RPY","short_pith_number":"pith:MS46TU7A","schema_version":"1.0","canonical_sha256":"64b9e9d3e09fdb62a6f21096b9f7b17e16a7827551878c06408ff8ca0b749524","source":{"kind":"arxiv","id":"2408.08812","version":2},"attestation_state":"computed","paper":{"title":"TRAM: Test-Time Risk Adaptation with Mixture of Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Amrit Singh Bedi, Amy Zhang, Hao Zhu, Mohamad Fares El Hajj Chehade","submitted_at":"2024-08-16T15:47:08Z","abstract_excerpt":"Deployed reinforcement learning agents often face safety requirements that are specified only after training, such as new hazard maps, revised risk thresholds, or behavioral alignment constraints. We study zero-update deployment-time adaptation, where a fixed library of risk-neutral source policies is reused under a newly specified reward-risk tradeoff. We propose TRAM (Test-Time Risk Adaptation via Mixture of Agents), a source-scored composition rule that evaluates each source policy under the target reward and an occupancy-based deployment risk, then selects actions using risk-adjusted sourc"},"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":"2408.08812","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-08-16T15:47:08Z","cross_cats_sorted":[],"title_canon_sha256":"74c38ce0cc5ba376e718df0efd304a483af9a326f4ad7ee62c89e829b5399e87","abstract_canon_sha256":"8396732ad5487868120f5b0c2cc45140a2d84f2857739732d46b89926ada2d34"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-21T01:04:11.934990Z","signature_b64":"NIe12LzW3NHF5D2VrvkrbLXnSrKrlQiXeoWJ462p0b3l2knX2THuj6XjS6cXAWbW5KK0P3Zx/ykhnvx8DMe9Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"64b9e9d3e09fdb62a6f21096b9f7b17e16a7827551878c06408ff8ca0b749524","last_reissued_at":"2026-05-21T01:04:11.934293Z","signature_status":"signed_v1","first_computed_at":"2026-05-21T01:04:11.934293Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TRAM: Test-Time Risk Adaptation with Mixture of Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Amrit Singh Bedi, Amy Zhang, Hao Zhu, Mohamad Fares El Hajj Chehade","submitted_at":"2024-08-16T15:47:08Z","abstract_excerpt":"Deployed reinforcement learning agents often face safety requirements that are specified only after training, such as new hazard maps, revised risk thresholds, or behavioral alignment constraints. We study zero-update deployment-time adaptation, where a fixed library of risk-neutral source policies is reused under a newly specified reward-risk tradeoff. We propose TRAM (Test-Time Risk Adaptation via Mixture of Agents), a source-scored composition rule that evaluates each source policy under the target reward and an occupancy-based deployment risk, then selects actions using risk-adjusted sourc"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.08812","kind":"arxiv","version":2},"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/2408.08812/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":"2408.08812","created_at":"2026-05-21T01:04:11.934390+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.08812v2","created_at":"2026-05-21T01:04:11.934390+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.08812","created_at":"2026-05-21T01:04:11.934390+00:00"},{"alias_kind":"pith_short_12","alias_value":"MS46TU7AT7NW","created_at":"2026-05-21T01:04:11.934390+00:00"},{"alias_kind":"pith_short_16","alias_value":"MS46TU7AT7NWFJXS","created_at":"2026-05-21T01:04:11.934390+00:00"},{"alias_kind":"pith_short_8","alias_value":"MS46TU7A","created_at":"2026-05-21T01:04:11.934390+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/MS46TU7AT7NWFJXSCCLLT55RPY","json":"https://pith.science/pith/MS46TU7AT7NWFJXSCCLLT55RPY.json","graph_json":"https://pith.science/api/pith-number/MS46TU7AT7NWFJXSCCLLT55RPY/graph.json","events_json":"https://pith.science/api/pith-number/MS46TU7AT7NWFJXSCCLLT55RPY/events.json","paper":"https://pith.science/paper/MS46TU7A"},"agent_actions":{"view_html":"https://pith.science/pith/MS46TU7AT7NWFJXSCCLLT55RPY","download_json":"https://pith.science/pith/MS46TU7AT7NWFJXSCCLLT55RPY.json","view_paper":"https://pith.science/paper/MS46TU7A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.08812&json=true","fetch_graph":"https://pith.science/api/pith-number/MS46TU7AT7NWFJXSCCLLT55RPY/graph.json","fetch_events":"https://pith.science/api/pith-number/MS46TU7AT7NWFJXSCCLLT55RPY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MS46TU7AT7NWFJXSCCLLT55RPY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MS46TU7AT7NWFJXSCCLLT55RPY/action/storage_attestation","attest_author":"https://pith.science/pith/MS46TU7AT7NWFJXSCCLLT55RPY/action/author_attestation","sign_citation":"https://pith.science/pith/MS46TU7AT7NWFJXSCCLLT55RPY/action/citation_signature","submit_replication":"https://pith.science/pith/MS46TU7AT7NWFJXSCCLLT55RPY/action/replication_record"}},"created_at":"2026-05-21T01:04:11.934390+00:00","updated_at":"2026-05-21T01:04:11.934390+00:00"}