{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DOABYQXNZQXX2ZV3JBT7MSB7QV","short_pith_number":"pith:DOABYQXN","schema_version":"1.0","canonical_sha256":"1b801c42edcc2f7d66bb4867f6483f85459277eddde846cc2f13b2271baa5399","source":{"kind":"arxiv","id":"2602.06575","version":2},"attestation_state":"computed","paper":{"title":"Think Proprioceptively: State-Grounded Visual Token Selection for VLA Policies","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.RO","authors_text":"Chengyang He, David Navarro-Alarcon, Fangyuan Wang, Guodong Guo, Jiaming Qi, Peng Zhou, Shipeng Lyu","submitted_at":"2026-02-06T10:16:22Z","abstract_excerpt":"Vision-language-action (VLA) models typically inject proprioception only as a late conditioning signal, preventing robot state from grounding instruction understanding or directing visual attention. We introduce ThinkProprio, which discretizes proprioception into VLM-vocabulary tokens and uses them jointly with the instruction to gate visual patches before VLM computation, steering the model toward action-relevant evidence while discarding redundant tokens early. We find that proprioception added as a passive conditioning signal leaves performance essentially unchanged; its value emerges when "},"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":"2602.06575","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-02-06T10:16:22Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"f5d6894ba5de3c610b9dcb868fe8994f6cafbfae5c730d5b03b841d5e0abb3ff","abstract_canon_sha256":"931fd7d71138b8351b726eab68217cb021873d410c6a3c27a0ed14d493ff3a06"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-07T02:19:46.756239Z","signature_b64":"cpGjOnfvZxn8OoqQ8HicEAkovNbOA9vLcGo0roGoOT2lJTcuHUa+xgg2myHYj/ZSiIS+5oFBtkzrmpb4c9JGDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1b801c42edcc2f7d66bb4867f6483f85459277eddde846cc2f13b2271baa5399","last_reissued_at":"2026-07-07T02:19:46.755343Z","signature_status":"signed_v1","first_computed_at":"2026-07-07T02:19:46.755343Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Think Proprioceptively: State-Grounded Visual Token Selection for VLA Policies","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.RO","authors_text":"Chengyang He, David Navarro-Alarcon, Fangyuan Wang, Guodong Guo, Jiaming Qi, Peng Zhou, Shipeng Lyu","submitted_at":"2026-02-06T10:16:22Z","abstract_excerpt":"Vision-language-action (VLA) models typically inject proprioception only as a late conditioning signal, preventing robot state from grounding instruction understanding or directing visual attention. We introduce ThinkProprio, which discretizes proprioception into VLM-vocabulary tokens and uses them jointly with the instruction to gate visual patches before VLM computation, steering the model toward action-relevant evidence while discarding redundant tokens early. We find that proprioception added as a passive conditioning signal leaves performance essentially unchanged; its value emerges when "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.06575","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/2602.06575/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":"2602.06575","created_at":"2026-07-07T02:19:46.755481+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.06575v2","created_at":"2026-07-07T02:19:46.755481+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.06575","created_at":"2026-07-07T02:19:46.755481+00:00"},{"alias_kind":"pith_short_12","alias_value":"DOABYQXNZQXX","created_at":"2026-07-07T02:19:46.755481+00:00"},{"alias_kind":"pith_short_16","alias_value":"DOABYQXNZQXX2ZV3","created_at":"2026-07-07T02:19:46.755481+00:00"},{"alias_kind":"pith_short_8","alias_value":"DOABYQXN","created_at":"2026-07-07T02:19:46.755481+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.08542","citing_title":"When Video Misreads: Closed-Loop Distillation of Reading Heuristics for Exploratory Manipulation Trace QA","ref_index":10,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DOABYQXNZQXX2ZV3JBT7MSB7QV","json":"https://pith.science/pith/DOABYQXNZQXX2ZV3JBT7MSB7QV.json","graph_json":"https://pith.science/api/pith-number/DOABYQXNZQXX2ZV3JBT7MSB7QV/graph.json","events_json":"https://pith.science/api/pith-number/DOABYQXNZQXX2ZV3JBT7MSB7QV/events.json","paper":"https://pith.science/paper/DOABYQXN"},"agent_actions":{"view_html":"https://pith.science/pith/DOABYQXNZQXX2ZV3JBT7MSB7QV","download_json":"https://pith.science/pith/DOABYQXNZQXX2ZV3JBT7MSB7QV.json","view_paper":"https://pith.science/paper/DOABYQXN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.06575&json=true","fetch_graph":"https://pith.science/api/pith-number/DOABYQXNZQXX2ZV3JBT7MSB7QV/graph.json","fetch_events":"https://pith.science/api/pith-number/DOABYQXNZQXX2ZV3JBT7MSB7QV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DOABYQXNZQXX2ZV3JBT7MSB7QV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DOABYQXNZQXX2ZV3JBT7MSB7QV/action/storage_attestation","attest_author":"https://pith.science/pith/DOABYQXNZQXX2ZV3JBT7MSB7QV/action/author_attestation","sign_citation":"https://pith.science/pith/DOABYQXNZQXX2ZV3JBT7MSB7QV/action/citation_signature","submit_replication":"https://pith.science/pith/DOABYQXNZQXX2ZV3JBT7MSB7QV/action/replication_record"}},"created_at":"2026-07-07T02:19:46.755481+00:00","updated_at":"2026-07-07T02:19:46.755481+00:00"}