{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ARR72KYUPLZDZRZV6KUHGTCHNH","short_pith_number":"pith:ARR72KYU","schema_version":"1.0","canonical_sha256":"0463fd2b147af23cc735f2a8734c4769dd8f723a225fa81ddaba517dbe122c18","source":{"kind":"arxiv","id":"2605.27079","version":1},"attestation_state":"computed","paper":{"title":"Trust Region Q Adjoint Matching","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.RO"],"primary_cat":"cs.LG","authors_text":"Changyeon Kim, Jaehyuk Kim, Jinwoo Shin, Kyungmin Lee, Yonghoon Dong","submitted_at":"2026-05-26T14:28:43Z","abstract_excerpt":"Off-policy reinforcement learning of pretrained flow policies remains challenging due to the instability of optimization arising from the multi-step sampling process. Recently, Q-learning with Adjoint Matching (QAM) addressed this issue by reformulating into a memoryless stochastic optimal control (SOC) problem with a learned critic. However, QAM inherits a fundamental fragility of critic-guided improvement: small critic errors are amplified when critics are ill-conditioned, often leading to model collapse. This paper introduces Trust Region Q-Adjoint Matching (TRQAM), a stable off-policy fine"},"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":"2605.27079","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-26T14:28:43Z","cross_cats_sorted":["cs.AI","cs.RO"],"title_canon_sha256":"215b7decb953b9745084fe54061a32feb4a5d18bd4e3d01d705ae57e33a99e62","abstract_canon_sha256":"bd026742da865e7984c4385363c9c8f331f45ca3ca4c763c5dface219a15c587"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T01:06:27.247491Z","signature_b64":"kBEiIro9BMNrWTgE6aJinFY2yTYTIwxdYmAvJ3i6ijlZFc4Wzz/5zKZtAMjOcNcAOdiUYMLoFQDkD+HLc1tbCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0463fd2b147af23cc735f2a8734c4769dd8f723a225fa81ddaba517dbe122c18","last_reissued_at":"2026-05-27T01:06:27.247025Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T01:06:27.247025Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Trust Region Q Adjoint Matching","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.RO"],"primary_cat":"cs.LG","authors_text":"Changyeon Kim, Jaehyuk Kim, Jinwoo Shin, Kyungmin Lee, Yonghoon Dong","submitted_at":"2026-05-26T14:28:43Z","abstract_excerpt":"Off-policy reinforcement learning of pretrained flow policies remains challenging due to the instability of optimization arising from the multi-step sampling process. Recently, Q-learning with Adjoint Matching (QAM) addressed this issue by reformulating into a memoryless stochastic optimal control (SOC) problem with a learned critic. However, QAM inherits a fundamental fragility of critic-guided improvement: small critic errors are amplified when critics are ill-conditioned, often leading to model collapse. This paper introduces Trust Region Q-Adjoint Matching (TRQAM), a stable off-policy fine"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.27079","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/2605.27079/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":"2605.27079","created_at":"2026-05-27T01:06:27.247101+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.27079v1","created_at":"2026-05-27T01:06:27.247101+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.27079","created_at":"2026-05-27T01:06:27.247101+00:00"},{"alias_kind":"pith_short_12","alias_value":"ARR72KYUPLZD","created_at":"2026-05-27T01:06:27.247101+00:00"},{"alias_kind":"pith_short_16","alias_value":"ARR72KYUPLZDZRZV","created_at":"2026-05-27T01:06:27.247101+00:00"},{"alias_kind":"pith_short_8","alias_value":"ARR72KYU","created_at":"2026-05-27T01:06:27.247101+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/ARR72KYUPLZDZRZV6KUHGTCHNH","json":"https://pith.science/pith/ARR72KYUPLZDZRZV6KUHGTCHNH.json","graph_json":"https://pith.science/api/pith-number/ARR72KYUPLZDZRZV6KUHGTCHNH/graph.json","events_json":"https://pith.science/api/pith-number/ARR72KYUPLZDZRZV6KUHGTCHNH/events.json","paper":"https://pith.science/paper/ARR72KYU"},"agent_actions":{"view_html":"https://pith.science/pith/ARR72KYUPLZDZRZV6KUHGTCHNH","download_json":"https://pith.science/pith/ARR72KYUPLZDZRZV6KUHGTCHNH.json","view_paper":"https://pith.science/paper/ARR72KYU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.27079&json=true","fetch_graph":"https://pith.science/api/pith-number/ARR72KYUPLZDZRZV6KUHGTCHNH/graph.json","fetch_events":"https://pith.science/api/pith-number/ARR72KYUPLZDZRZV6KUHGTCHNH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ARR72KYUPLZDZRZV6KUHGTCHNH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ARR72KYUPLZDZRZV6KUHGTCHNH/action/storage_attestation","attest_author":"https://pith.science/pith/ARR72KYUPLZDZRZV6KUHGTCHNH/action/author_attestation","sign_citation":"https://pith.science/pith/ARR72KYUPLZDZRZV6KUHGTCHNH/action/citation_signature","submit_replication":"https://pith.science/pith/ARR72KYUPLZDZRZV6KUHGTCHNH/action/replication_record"}},"created_at":"2026-05-27T01:06:27.247101+00:00","updated_at":"2026-05-27T01:06:27.247101+00:00"}