{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:RE2WLRMIB6W73PK2LHATYHJB7O","short_pith_number":"pith:RE2WLRMI","canonical_record":{"source":{"id":"2605.15157","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-05-14T17:51:40Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"c048f0a39f28d3aff1857ff1755136452a5c20b4481281a8372b6ab987fe90a5","abstract_canon_sha256":"8adf952441590f81f4c6b6d6af7f722a24f6bdc18c5e0b873c219eb7f4c2ec3e"},"schema_version":"1.0"},"canonical_sha256":"893565c5880fadfdbd5a59c13c1d21fb908f22c45575bdb2b3b852d144fe4054","source":{"kind":"arxiv","id":"2605.15157","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.15157","created_at":"2026-05-17T21:18:33Z"},{"alias_kind":"arxiv_version","alias_value":"2605.15157v1","created_at":"2026-05-17T21:18:33Z"},{"alias_kind":"pith_short_12","alias_value":"RE2WLRMIB6W7","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"RE2WLRMIB6W73PK2","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"RE2WLRMI","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:RE2WLRMIB6W73PK2LHATYHJB7O","target":"record","payload":{"canonical_record":{"source":{"id":"2605.15157","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-05-14T17:51:40Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"c048f0a39f28d3aff1857ff1755136452a5c20b4481281a8372b6ab987fe90a5","abstract_canon_sha256":"8adf952441590f81f4c6b6d6af7f722a24f6bdc18c5e0b873c219eb7f4c2ec3e"},"schema_version":"1.0"},"canonical_sha256":"893565c5880fadfdbd5a59c13c1d21fb908f22c45575bdb2b3b852d144fe4054","receipt":{"kind":"pith_receipt","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.2","canonical_sha256":"893565c5880fadfdbd5a59c13c1d21fb908f22c45575bdb2b3b852d144fe4054","last_reissued_at":"2026-05-17T21:57:18.749021Z","signature_status":"unsigned_v0","first_computed_at":"2026-05-17T21:40:25.426639Z"},"source_kind":"arxiv","source_id":"2605.15157","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T21:18:33Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"eD/s3r2YhBLrkIivrkYtfzR1X0Q6MK/CavvIiI0LN7mLLahwVl+SwlwwXJj995y/kl5Gdt3vVRMJ1WAiw0RwCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-25T23:09:58.212278Z"},"content_sha256":"1626721495e168f3fa4b5b57e8996b4caa950245aade65007c6106d62f757785","schema_version":"1.0","event_id":"sha256:1626721495e168f3fa4b5b57e8996b4caa950245aade65007c6106d62f757785"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:RE2WLRMIB6W73PK2LHATYHJB7O","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Hand-in-the-Loop: Improving Dexterous VLA via Seamless Interventional Correction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"HandITL blends human corrective intent with ongoing VLA policy execution to eliminate gesture jumps during dexterous hand takeovers.","cross_cats":["cs.LG"],"primary_cat":"cs.RO","authors_text":"Liqun Huang, Nie Lin, Ruoshi Wen, Wei Xu, Xiao Ma, Xinjun Sheng, Zhengming Zhu, Zhuohang Li","submitted_at":"2026-05-14T17:51:40Z","abstract_excerpt":"Vision-Language-Action (VLA) models are prone to compounding errors in dexterous manipulation, where high-dimensional action spaces and contact-rich dynamics amplify small policy deviations over long horizons. While Interactive Imitation Learning (IIL) can refine policies through human takeover data, applying it to high-degree-of-freedom (DoF) robotic hands remains challenging due to a command mismatch between human teleoperation and policy execution at the takeover moment, which causes abrupt robot-hand configuration changes, or \"gesture jumps\". We present Hand-in-the-Loop (HandITL), a seamle"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"HandITL reduces takeover jitter by 99.8% and grasp failures by 87.5%, mean completion time by 19.1%, and produces policies that outperform those trained with standard teleoperation data by 19% on average across three long-horizon dexterous tasks.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the seamless blending of human corrective intent with autonomous policy execution can be achieved without introducing new instabilities or losing the benefit of the human correction in high-dimensional action spaces and contact-rich dynamics.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"HandITL blends human intent with policy execution to eliminate gesture jumps in dexterous VLA interventions, cutting jitter by 99.8%, grasp failures by 87.5%, and yielding 19% better refined policies.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"HandITL blends human corrective intent with ongoing VLA policy execution to eliminate gesture jumps during dexterous hand takeovers.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ec41bb2ceeee400f81d027d6c85984508cf4993011ae5ece556bdfd02f3c2099"},"source":{"id":"2605.15157","kind":"arxiv","version":1},"verdict":{"id":"acab7a65-81fe-4604-9b9a-4c59c6f5117d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T03:05:35.135590Z","strongest_claim":"HandITL reduces takeover jitter by 99.8% and grasp failures by 87.5%, mean completion time by 19.1%, and produces policies that outperform those trained with standard teleoperation data by 19% on average across three long-horizon dexterous tasks.","one_line_summary":"HandITL blends human intent with policy execution to eliminate gesture jumps in dexterous VLA interventions, cutting jitter by 99.8%, grasp failures by 87.5%, and yielding 19% better refined policies.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the seamless blending of human corrective intent with autonomous policy execution can be achieved without introducing new instabilities or losing the benefit of the human correction in high-dimensional action spaces and contact-rich dynamics.","pith_extraction_headline":"HandITL blends human corrective intent with ongoing VLA policy execution to eliminate gesture jumps during dexterous hand takeovers."},"references":{"count":34,"sample":[{"doi":"","year":2019,"title":"Sample efficient interactive end-to-end deep learning for self-driving cars with selective multi-class safe dataset aggregation","work_id":"acd7b7a4-c530-4a71-b0c8-784b3b16428b","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"GR00T N1: An Open Foundation Model for Generalist Humanoid Robots","work_id":"e2db69c7-ee8a-4cb7-a761-7b8de1dfcf97","ref_index":2,"cited_arxiv_id":"2503.14734","is_internal_anchor":true},{"doi":"","year":2024,"title":"$\\pi_0$: A Vision-Language-Action Flow Model for General Robot Control","work_id":"f790abdc-a796-482f-a40d-f8ee035ecfc2","ref_index":3,"cited_arxiv_id":"2410.24164","is_internal_anchor":true},{"doi":"","year":2025,"title":"GR-3 Technical Report","work_id":"7f3b800b-0146-4c37-b178-0801cd4270db","ref_index":4,"cited_arxiv_id":"2507.15493","is_internal_anchor":true},{"doi":"","year":2025,"title":"Conrft: A reinforced fine-tuning method for vla models via consistency policy","work_id":"12e4ef8e-5311-42c5-86de-e8a0c0a4c1c1","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":34,"snapshot_sha256":"7413377d38a8bd29f190194c953a2e264ba0fd7cf5e887b4652ad54dbd1982e8","internal_anchors":6},"formal_canon":{"evidence_count":2,"snapshot_sha256":"c5c069f49630d5d70a4d6eaab3e61864c2310c7360dbc4bfbe32bb29f1378c60"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"acab7a65-81fe-4604-9b9a-4c59c6f5117d"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T21:57:18Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"GWFk9msBDTBp9PAmMjb0ZPf1MTx3MKb2FdVBxSrWsmBuHqCc1HXVjp74M4TLfKuxcOWiYeST0XU+Na3qEjWDAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-25T23:09:58.213024Z"},"content_sha256":"5c48f5dfc42994bb0d37d4e5547ab67d9609418bcf48f29e359d4ce711f9cf94","schema_version":"1.0","event_id":"sha256:5c48f5dfc42994bb0d37d4e5547ab67d9609418bcf48f29e359d4ce711f9cf94"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/RE2WLRMIB6W73PK2LHATYHJB7O/bundle.json","state_url":"https://pith.science/pith/RE2WLRMIB6W73PK2LHATYHJB7O/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/RE2WLRMIB6W73PK2LHATYHJB7O/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":"2026-05-25T23:09:58Z","links":{"resolver":"https://pith.science/pith/RE2WLRMIB6W73PK2LHATYHJB7O","bundle":"https://pith.science/pith/RE2WLRMIB6W73PK2LHATYHJB7O/bundle.json","state":"https://pith.science/pith/RE2WLRMIB6W73PK2LHATYHJB7O/state.json","well_known_bundle":"https://pith.science/.well-known/pith/RE2WLRMIB6W73PK2LHATYHJB7O/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:RE2WLRMIB6W73PK2LHATYHJB7O","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":"8adf952441590f81f4c6b6d6af7f722a24f6bdc18c5e0b873c219eb7f4c2ec3e","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-05-14T17:51:40Z","title_canon_sha256":"c048f0a39f28d3aff1857ff1755136452a5c20b4481281a8372b6ab987fe90a5"},"schema_version":"1.0","source":{"id":"2605.15157","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.15157","created_at":"2026-05-17T21:18:33Z"},{"alias_kind":"arxiv_version","alias_value":"2605.15157v1","created_at":"2026-05-17T21:18:33Z"},{"alias_kind":"pith_short_12","alias_value":"RE2WLRMIB6W7","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"RE2WLRMIB6W73PK2","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"RE2WLRMI","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:5c48f5dfc42994bb0d37d4e5547ab67d9609418bcf48f29e359d4ce711f9cf94","target":"graph","created_at":"2026-05-17T21:57:18Z","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":"HandITL reduces takeover jitter by 99.8% and grasp failures by 87.5%, mean completion time by 19.1%, and produces policies that outperform those trained with standard teleoperation data by 19% on average across three long-horizon dexterous tasks."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the seamless blending of human corrective intent with autonomous policy execution can be achieved without introducing new instabilities or losing the benefit of the human correction in high-dimensional action spaces and contact-rich dynamics."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"HandITL blends human intent with policy execution to eliminate gesture jumps in dexterous VLA interventions, cutting jitter by 99.8%, grasp failures by 87.5%, and yielding 19% better refined policies."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"HandITL blends human corrective intent with ongoing VLA policy execution to eliminate gesture jumps during dexterous hand takeovers."}],"snapshot_sha256":"ec41bb2ceeee400f81d027d6c85984508cf4993011ae5ece556bdfd02f3c2099"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"c5c069f49630d5d70a4d6eaab3e61864c2310c7360dbc4bfbe32bb29f1378c60"},"paper":{"abstract_excerpt":"Vision-Language-Action (VLA) models are prone to compounding errors in dexterous manipulation, where high-dimensional action spaces and contact-rich dynamics amplify small policy deviations over long horizons. While Interactive Imitation Learning (IIL) can refine policies through human takeover data, applying it to high-degree-of-freedom (DoF) robotic hands remains challenging due to a command mismatch between human teleoperation and policy execution at the takeover moment, which causes abrupt robot-hand configuration changes, or \"gesture jumps\". We present Hand-in-the-Loop (HandITL), a seamle","authors_text":"Liqun Huang, Nie Lin, Ruoshi Wen, Wei Xu, Xiao Ma, Xinjun Sheng, Zhengming Zhu, Zhuohang Li","cross_cats":["cs.LG"],"headline":"HandITL blends human corrective intent with ongoing VLA policy execution to eliminate gesture jumps during dexterous hand takeovers.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-05-14T17:51:40Z","title":"Hand-in-the-Loop: Improving Dexterous VLA via Seamless Interventional Correction"},"references":{"count":34,"internal_anchors":6,"resolved_work":34,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Sample efficient interactive end-to-end deep learning for self-driving cars with selective multi-class safe dataset aggregation","work_id":"acd7b7a4-c530-4a71-b0c8-784b3b16428b","year":2019},{"cited_arxiv_id":"2503.14734","doi":"","is_internal_anchor":true,"ref_index":2,"title":"GR00T N1: An Open Foundation Model for Generalist Humanoid Robots","work_id":"e2db69c7-ee8a-4cb7-a761-7b8de1dfcf97","year":2025},{"cited_arxiv_id":"2410.24164","doi":"","is_internal_anchor":true,"ref_index":3,"title":"$\\pi_0$: A Vision-Language-Action Flow Model for General Robot Control","work_id":"f790abdc-a796-482f-a40d-f8ee035ecfc2","year":2024},{"cited_arxiv_id":"2507.15493","doi":"","is_internal_anchor":true,"ref_index":4,"title":"GR-3 Technical Report","work_id":"7f3b800b-0146-4c37-b178-0801cd4270db","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Conrft: A reinforced fine-tuning method for vla models via consistency policy","work_id":"12e4ef8e-5311-42c5-86de-e8a0c0a4c1c1","year":2025}],"snapshot_sha256":"7413377d38a8bd29f190194c953a2e264ba0fd7cf5e887b4652ad54dbd1982e8"},"source":{"id":"2605.15157","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-15T03:05:35.135590Z","id":"acab7a65-81fe-4604-9b9a-4c59c6f5117d","model_set":{"reader":"grok-4.3"},"one_line_summary":"HandITL blends human intent with policy execution to eliminate gesture jumps in dexterous VLA interventions, cutting jitter by 99.8%, grasp failures by 87.5%, and yielding 19% better refined policies.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"HandITL blends human corrective intent with ongoing VLA policy execution to eliminate gesture jumps during dexterous hand takeovers.","strongest_claim":"HandITL reduces takeover jitter by 99.8% and grasp failures by 87.5%, mean completion time by 19.1%, and produces policies that outperform those trained with standard teleoperation data by 19% on average across three long-horizon dexterous tasks.","weakest_assumption":"That the seamless blending of human corrective intent with autonomous policy execution can be achieved without introducing new instabilities or losing the benefit of the human correction in high-dimensional action spaces and contact-rich dynamics."}},"verdict_id":"acab7a65-81fe-4604-9b9a-4c59c6f5117d"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:1626721495e168f3fa4b5b57e8996b4caa950245aade65007c6106d62f757785","target":"record","created_at":"2026-05-17T21:18:33Z","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":"8adf952441590f81f4c6b6d6af7f722a24f6bdc18c5e0b873c219eb7f4c2ec3e","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-05-14T17:51:40Z","title_canon_sha256":"c048f0a39f28d3aff1857ff1755136452a5c20b4481281a8372b6ab987fe90a5"},"schema_version":"1.0","source":{"id":"2605.15157","kind":"arxiv","version":1}},"canonical_sha256":"893565c5880fadfdbd5a59c13c1d21fb908f22c45575bdb2b3b852d144fe4054","receipt":{"builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"893565c5880fadfdbd5a59c13c1d21fb908f22c45575bdb2b3b852d144fe4054","first_computed_at":"2026-05-17T21:40:25.426639Z","kind":"pith_receipt","last_reissued_at":"2026-05-17T21:57:18.749021Z","receipt_version":"0.2","signature_status":"unsigned_v0"},"source_id":"2605.15157","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:1626721495e168f3fa4b5b57e8996b4caa950245aade65007c6106d62f757785","sha256:5c48f5dfc42994bb0d37d4e5547ab67d9609418bcf48f29e359d4ce711f9cf94"],"state_sha256":"5ae06dbb12f083b84e58a0d917b8f66f286f973507b8385dec73f2a7616228e1"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"DEq9YTkybtQ+tjUohwWeU8gurOAd9T4m5raAlchX+5Rya50L2Wa70+nYFPWr1ai5wGLb4x3CBKG9sZMC/tPiAQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-25T23:09:58.216793Z","bundle_sha256":"b79860735086d57ccc147d21695fa6f82a355cae4fc7adf3954a2ac3a4d30eb4"}}