{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:N4IZL755I5DOJXZC62N47S5WCK","short_pith_number":"pith:N4IZL755","canonical_record":{"source":{"id":"2605.05172","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-05-06T17:40:11Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2289b6c1cb3b844a53215203c2a9025eea500d5977c638482e256a364f3c18ef","abstract_canon_sha256":"a2ed216db8a185cc88bfa44f57ced640568daf8b3298f69501ba63de39c8c156"},"schema_version":"1.0"},"canonical_sha256":"6f1195ffbd4746e4df22f69bcfcbb612a1c854e911649505472ddd6e45169407","source":{"kind":"arxiv","id":"2605.05172","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.05172","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"arxiv_version","alias_value":"2605.05172v2","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.05172","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"pith_short_12","alias_value":"N4IZL755I5DO","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"pith_short_16","alias_value":"N4IZL755I5DOJXZC","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"pith_short_8","alias_value":"N4IZL755","created_at":"2026-06-19T16:09:58Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:N4IZL755I5DOJXZC62N47S5WCK","target":"record","payload":{"canonical_record":{"source":{"id":"2605.05172","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-05-06T17:40:11Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2289b6c1cb3b844a53215203c2a9025eea500d5977c638482e256a364f3c18ef","abstract_canon_sha256":"a2ed216db8a185cc88bfa44f57ced640568daf8b3298f69501ba63de39c8c156"},"schema_version":"1.0"},"canonical_sha256":"6f1195ffbd4746e4df22f69bcfcbb612a1c854e911649505472ddd6e45169407","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:09:58.965700Z","signature_b64":"oGUJs930jJKMm5XAi8IB7PWM2e8qgkY6q+o4BgeoLb7hyHC5UgbSMAlXAL0iZQVk6S2ofV0hHMhMfxLBk/f3CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6f1195ffbd4746e4df22f69bcfcbb612a1c854e911649505472ddd6e45169407","last_reissued_at":"2026-06-19T16:09:58.965344Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:09:58.965344Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.05172","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-19T16:09:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/dCW1YVke4u1gyr7Dpxg48idFOqPeKN0XnWyCEgfMp4fsOjbU/INY/R3xf/xJCywVocKceVgMn0ggibFA7MLBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T04:12:54.217706Z"},"content_sha256":"082371ffc262f92d7d48be5b4a0caab15a5efce6f6b01362f2e466b8693515d1","schema_version":"1.0","event_id":"sha256:082371ffc262f92d7d48be5b4a0caab15a5efce6f6b01362f2e466b8693515d1"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:N4IZL755I5DOJXZC62N47S5WCK","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"When Life Gives You BC, Make Q-functions: Extracting Q-values from Behavior Cloning for On-Robot Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Extracting Q-values from a behavior cloning policy lets robots improve online with reinforcement learning while avoiding the usual distribution mismatch.","cross_cats":["cs.AI"],"primary_cat":"cs.RO","authors_text":"Karl Schmeckpeper, Lakshita Dodeja, Ondrej Biza, Robin Walters, Shivam Vats, Stefanie Tellex, Stephen Hart, Thomas Weng","submitted_at":"2026-05-06T17:40:11Z","abstract_excerpt":"Behavior Cloning (BC) has emerged as a highly effective paradigm for robot learning. However, BC lacks a self-guided mechanism for online improvement after demonstrations have been collected. Existing offline-to-online learning methods often cause policies to replace previously learned good actions due to a distribution mismatch between offline data and online learning. In this work, we propose Q2RL, Q-Estimation and Q-Gating from BC for Reinforcement Learning, an algorithm for efficient offline-to-online learning. Our method consists of two parts: (1) Q-Estimation extracts a Q-function from a"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Q2RL outperforms SOTA offline-to-online learning baselines on success rate and time to convergence. Q2RL is efficient enough to be applied in an on-robot RL setting, learning robust policies for contact-rich and high precision manipulation tasks such as pipe assembly and kitting, in 1-2 hours of online interaction, achieving success rates of up to 100% and up to 3.75x improvement against the original BC policy.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The Q-function extracted from the BC policy with only a few interactions is sufficiently accurate to guide gating and RL training without reintroducing distribution mismatch or value estimation errors.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Q2RL extracts Q-functions from BC policies via minimal interactions and applies Q-gating to enable stable offline-to-online RL, outperforming baselines on manipulation benchmarks and achieving up to 100% success on-robot.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Extracting Q-values from a behavior cloning policy lets robots improve online with reinforcement learning while avoiding the usual distribution mismatch.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"33dddf0fc0b10d631df4c5f54321afb28ee3c2842b23987185994d7473c7e4d0"},"source":{"id":"2605.05172","kind":"arxiv","version":2},"verdict":{"id":"5b754f63-ac5e-4cb7-bba1-e90bdff137a1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T16:29:17.909019Z","strongest_claim":"Q2RL outperforms SOTA offline-to-online learning baselines on success rate and time to convergence. Q2RL is efficient enough to be applied in an on-robot RL setting, learning robust policies for contact-rich and high precision manipulation tasks such as pipe assembly and kitting, in 1-2 hours of online interaction, achieving success rates of up to 100% and up to 3.75x improvement against the original BC policy.","one_line_summary":"Q2RL extracts Q-functions from BC policies via minimal interactions and applies Q-gating to enable stable offline-to-online RL, outperforming baselines on manipulation benchmarks and achieving up to 100% success on-robot.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The Q-function extracted from the BC policy with only a few interactions is sufficiently accurate to guide gating and RL training without reintroducing distribution mismatch or value estimation errors.","pith_extraction_headline":"Extracting Q-values from a behavior cloning policy lets robots improve online with reinforcement learning while avoiding the usual distribution mismatch."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.05172/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T10:35:46.506270Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T21:01:19.771120Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T13:47:26.009830Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"061b1eb5af48ba6f1076f24a7f993d2ad1f213fc1e78e95b26c48119d695a87c"},"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"},"verdict_id":"5b754f63-ac5e-4cb7-bba1-e90bdff137a1"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-19T16:09:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"oQ/yKWYmQa4FajpdOXNUNHLv6pGLIEFtghfCSeXFs0wp7vMvxHnN2iLFf1IkWBMDYt+GcG4KjEe5Z1YYr0rJDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T04:12:54.218478Z"},"content_sha256":"1c170449f252dd0998cd52586e82c0f9e98106a9f5acea522f32de6c7830d2bc","schema_version":"1.0","event_id":"sha256:1c170449f252dd0998cd52586e82c0f9e98106a9f5acea522f32de6c7830d2bc"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/N4IZL755I5DOJXZC62N47S5WCK/bundle.json","state_url":"https://pith.science/pith/N4IZL755I5DOJXZC62N47S5WCK/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/N4IZL755I5DOJXZC62N47S5WCK/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-06T04:12:54Z","links":{"resolver":"https://pith.science/pith/N4IZL755I5DOJXZC62N47S5WCK","bundle":"https://pith.science/pith/N4IZL755I5DOJXZC62N47S5WCK/bundle.json","state":"https://pith.science/pith/N4IZL755I5DOJXZC62N47S5WCK/state.json","well_known_bundle":"https://pith.science/.well-known/pith/N4IZL755I5DOJXZC62N47S5WCK/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:N4IZL755I5DOJXZC62N47S5WCK","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":"a2ed216db8a185cc88bfa44f57ced640568daf8b3298f69501ba63de39c8c156","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-05-06T17:40:11Z","title_canon_sha256":"2289b6c1cb3b844a53215203c2a9025eea500d5977c638482e256a364f3c18ef"},"schema_version":"1.0","source":{"id":"2605.05172","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.05172","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"arxiv_version","alias_value":"2605.05172v2","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.05172","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"pith_short_12","alias_value":"N4IZL755I5DO","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"pith_short_16","alias_value":"N4IZL755I5DOJXZC","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"pith_short_8","alias_value":"N4IZL755","created_at":"2026-06-19T16:09:58Z"}],"graph_snapshots":[{"event_id":"sha256:1c170449f252dd0998cd52586e82c0f9e98106a9f5acea522f32de6c7830d2bc","target":"graph","created_at":"2026-06-19T16:09:58Z","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":"Q2RL outperforms SOTA offline-to-online learning baselines on success rate and time to convergence. Q2RL is efficient enough to be applied in an on-robot RL setting, learning robust policies for contact-rich and high precision manipulation tasks such as pipe assembly and kitting, in 1-2 hours of online interaction, achieving success rates of up to 100% and up to 3.75x improvement against the original BC policy."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The Q-function extracted from the BC policy with only a few interactions is sufficiently accurate to guide gating and RL training without reintroducing distribution mismatch or value estimation errors."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Q2RL extracts Q-functions from BC policies via minimal interactions and applies Q-gating to enable stable offline-to-online RL, outperforming baselines on manipulation benchmarks and achieving up to 100% success on-robot."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Extracting Q-values from a behavior cloning policy lets robots improve online with reinforcement learning while avoiding the usual distribution mismatch."}],"snapshot_sha256":"33dddf0fc0b10d631df4c5f54321afb28ee3c2842b23987185994d7473c7e4d0"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T10:35:46.506270Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T21:01:19.771120Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T13:47:26.009830Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.05172/integrity.json","findings":[],"snapshot_sha256":"061b1eb5af48ba6f1076f24a7f993d2ad1f213fc1e78e95b26c48119d695a87c","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Behavior Cloning (BC) has emerged as a highly effective paradigm for robot learning. However, BC lacks a self-guided mechanism for online improvement after demonstrations have been collected. Existing offline-to-online learning methods often cause policies to replace previously learned good actions due to a distribution mismatch between offline data and online learning. In this work, we propose Q2RL, Q-Estimation and Q-Gating from BC for Reinforcement Learning, an algorithm for efficient offline-to-online learning. Our method consists of two parts: (1) Q-Estimation extracts a Q-function from a","authors_text":"Karl Schmeckpeper, Lakshita Dodeja, Ondrej Biza, Robin Walters, Shivam Vats, Stefanie Tellex, Stephen Hart, Thomas Weng","cross_cats":["cs.AI"],"headline":"Extracting Q-values from a behavior cloning policy lets robots improve online with reinforcement learning while avoiding the usual distribution mismatch.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-05-06T17:40:11Z","title":"When Life Gives You BC, Make Q-functions: Extracting Q-values from Behavior Cloning for On-Robot Reinforcement Learning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.05172","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-08T16:29:17.909019Z","id":"5b754f63-ac5e-4cb7-bba1-e90bdff137a1","model_set":{"reader":"grok-4.3"},"one_line_summary":"Q2RL extracts Q-functions from BC policies via minimal interactions and applies Q-gating to enable stable offline-to-online RL, outperforming baselines on manipulation benchmarks and achieving up to 100% success on-robot.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Extracting Q-values from a behavior cloning policy lets robots improve online with reinforcement learning while avoiding the usual distribution mismatch.","strongest_claim":"Q2RL outperforms SOTA offline-to-online learning baselines on success rate and time to convergence. Q2RL is efficient enough to be applied in an on-robot RL setting, learning robust policies for contact-rich and high precision manipulation tasks such as pipe assembly and kitting, in 1-2 hours of online interaction, achieving success rates of up to 100% and up to 3.75x improvement against the original BC policy.","weakest_assumption":"The Q-function extracted from the BC policy with only a few interactions is sufficiently accurate to guide gating and RL training without reintroducing distribution mismatch or value estimation errors."}},"verdict_id":"5b754f63-ac5e-4cb7-bba1-e90bdff137a1"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:082371ffc262f92d7d48be5b4a0caab15a5efce6f6b01362f2e466b8693515d1","target":"record","created_at":"2026-06-19T16:09:58Z","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":"a2ed216db8a185cc88bfa44f57ced640568daf8b3298f69501ba63de39c8c156","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-05-06T17:40:11Z","title_canon_sha256":"2289b6c1cb3b844a53215203c2a9025eea500d5977c638482e256a364f3c18ef"},"schema_version":"1.0","source":{"id":"2605.05172","kind":"arxiv","version":2}},"canonical_sha256":"6f1195ffbd4746e4df22f69bcfcbb612a1c854e911649505472ddd6e45169407","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"6f1195ffbd4746e4df22f69bcfcbb612a1c854e911649505472ddd6e45169407","first_computed_at":"2026-06-19T16:09:58.965344Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-19T16:09:58.965344Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"oGUJs930jJKMm5XAi8IB7PWM2e8qgkY6q+o4BgeoLb7hyHC5UgbSMAlXAL0iZQVk6S2ofV0hHMhMfxLBk/f3CA==","signature_status":"signed_v1","signed_at":"2026-06-19T16:09:58.965700Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.05172","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:082371ffc262f92d7d48be5b4a0caab15a5efce6f6b01362f2e466b8693515d1","sha256:1c170449f252dd0998cd52586e82c0f9e98106a9f5acea522f32de6c7830d2bc"],"state_sha256":"74de8cfba61fc476802af16eb697935589808a218cd82dae619833992b4d26b5"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"aOYdaSfl/WcQfdLNHfkABh0bpG1HhWb8ud8tuIrgysAGaHIsSphj0hOh1F7Kj8RCIthS13i74lYmFFOmKHskCg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-06T04:12:54.222647Z","bundle_sha256":"0ce7e273d074d4ecbe9990a5a0cef39ff070eddc2850972daf74f0ec47be0c39"}}