{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:2YNZJQXKERHE44UGTNN3W6KL3R","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":"c5b4a4bb45cbd5cc2bcec4f144608d501ff9ef6fb5eb15de46ef0744c9be4430","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-05T11:29:17Z","title_canon_sha256":"70cb4e591b6b3f80a2eb6f4194501f2e5ff9d1017e8ad8e834565b90df3e761d"},"schema_version":"1.0","source":{"id":"2603.05066","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2603.05066","created_at":"2026-05-20T01:06:08Z"},{"alias_kind":"arxiv_version","alias_value":"2603.05066v3","created_at":"2026-05-20T01:06:08Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.05066","created_at":"2026-05-20T01:06:08Z"},{"alias_kind":"pith_short_12","alias_value":"2YNZJQXKERHE","created_at":"2026-05-20T01:06:08Z"},{"alias_kind":"pith_short_16","alias_value":"2YNZJQXKERHE44UG","created_at":"2026-05-20T01:06:08Z"},{"alias_kind":"pith_short_8","alias_value":"2YNZJQXK","created_at":"2026-05-20T01:06:08Z"}],"graph_snapshots":[{"event_id":"sha256:dea81bbb99de22a70e705dd00caae66c36f1639e68f35e1198a57a9702e5ded1","target":"graph","created_at":"2026-05-20T01:06:08Z","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":"RCRL improves sample efficiency under the nominal reward parameterization, enables efficient adaptation to new parameterizations, and supports zero-shot behavioral adjustment at deployment."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That recomputing counterfactual rewards from replay data collected under the nominal policy produces unbiased training signals for other reward parameterizations."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"RCRL conditions RL policies on reward parameters and uses shared replay data to train for multiple objectives under a single nominal reward, improving efficiency and adaptability."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Conditioning RL agents on reward parameters during single-objective training enables zero-shot adaptation to new rewards via replay data alone."}],"snapshot_sha256":"ec667cc40c61ec88d2bb4450fb4e3a40c40f419c5ecff6bbb8650ca19c74c784"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"ef1afa9b6df446ae10a438a688321c081a7446c6e8340ec9c70cb8bf942e7a4e"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2603.05066/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Single-task RL agents are typically trained under a fixed reward function, which limits their robustness to reward misspecification and their ability to adapt to changing preferences. We introduce Reward-Conditioned Reinforcement Learning (RCRL), an off-policy method that conditions agents on reward parameterizations while collecting experience under a single nominal objective. By recomputing counterfactual rewards from shared replay data, RCRL exposes the agent to multiple reward objectives without additional environment interaction, connecting single-task RL with ideas from multi-objective a","authors_text":"Marek Cygan, Michal Nauman, Pieter Abbeel","cross_cats":[],"headline":"Conditioning RL agents on reward parameters during single-objective training enables zero-shot adaptation to new rewards via replay data alone.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-05T11:29:17Z","title":"Reward-Conditioned Reinforcement Learning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.05066","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-15T16:05:50.794154Z","id":"abd32904-dc8c-45a7-ab96-799f4e4d18f8","model_set":{"reader":"grok-4.3"},"one_line_summary":"RCRL conditions RL policies on reward parameters and uses shared replay data to train for multiple objectives under a single nominal reward, improving efficiency and adaptability.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Conditioning RL agents on reward parameters during single-objective training enables zero-shot adaptation to new rewards via replay data alone.","strongest_claim":"RCRL improves sample efficiency under the nominal reward parameterization, enables efficient adaptation to new parameterizations, and supports zero-shot behavioral adjustment at deployment.","weakest_assumption":"That recomputing counterfactual rewards from replay data collected under the nominal policy produces unbiased training signals for other reward parameterizations."}},"verdict_id":"abd32904-dc8c-45a7-ab96-799f4e4d18f8"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:7e14a18f41733eaa9eb1a7c8afe783cdeed7be0135f56b6a55ebb4bd9d4af34f","target":"record","created_at":"2026-05-20T01:06:08Z","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":"c5b4a4bb45cbd5cc2bcec4f144608d501ff9ef6fb5eb15de46ef0744c9be4430","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-05T11:29:17Z","title_canon_sha256":"70cb4e591b6b3f80a2eb6f4194501f2e5ff9d1017e8ad8e834565b90df3e761d"},"schema_version":"1.0","source":{"id":"2603.05066","kind":"arxiv","version":3}},"canonical_sha256":"d61b94c2ea244e4e72869b5bbb794bdc6a3c7c3d2353b502e82050d6f4510a50","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d61b94c2ea244e4e72869b5bbb794bdc6a3c7c3d2353b502e82050d6f4510a50","first_computed_at":"2026-05-20T01:06:08.893422Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T01:06:08.893422Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"XU8XaRUzEfBrp0dERnYGULt1RzB5bG7ZSyGtADM9dPyD6ndM4/aaBMpeIddVSWhCyfbZxqhLao+0Q0b2lT7jBA==","signature_status":"signed_v1","signed_at":"2026-05-20T01:06:08.894232Z","signed_message":"canonical_sha256_bytes"},"source_id":"2603.05066","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:7e14a18f41733eaa9eb1a7c8afe783cdeed7be0135f56b6a55ebb4bd9d4af34f","sha256:dea81bbb99de22a70e705dd00caae66c36f1639e68f35e1198a57a9702e5ded1"],"state_sha256":"994ab29060e84dc7aa740e57a2ccd8ac51ceb625b65eda03b490aac330e0272a"}