{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:6OAZJRETLC6UF4K7R7RRKRNQPW","short_pith_number":"pith:6OAZJRET","schema_version":"1.0","canonical_sha256":"f38194c49358bd42f15f8fe31545b07d848893f416046fc463029a7679cc3dc4","source":{"kind":"arxiv","id":"2606.10305","version":1},"attestation_state":"computed","paper":{"title":"SARM2: Multi-Task Stage Aware Reward Modeling for Self Improving Robotic Manipulation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Chuan Wen, Hau Zheng, Jiankai Sun, Justin Yu, Ken Goldberg, Mac Schwager, Philipp Wu, Pieter Abbeel, Qianzhong Chen, Suning Huang, Yide Shentu","submitted_at":"2026-06-09T01:46:23Z","abstract_excerpt":"Fine-tuning vision-language-action (VLA) policies for long-horizon manipulation still relies heavily on behavior cloning, which requires costly high-quality demonstrations and keeps policies near the demonstration distribution. Reward models can reduce this dependence by reweighting demonstrations and providing dense supervision for on-robot reinforcement learning (RL), but they must be dense, accurate, and general. Existing methods fall short: task-specific stage-aware models are accurate but require per-task annotations, while general vision-language-model (VLM) reward models are broadly app"},"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":"2606.10305","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-06-09T01:46:23Z","cross_cats_sorted":[],"title_canon_sha256":"7006ec12c8b7e79be9e34c97e85ada602c92645d0914deeaac460b5d14bedaa1","abstract_canon_sha256":"95236a3280522cde2aca4f80c46f032615a6737605ae1460e9b267f42eae72c6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-10T01:10:10.859567Z","signature_b64":"49jXmYF65oSTLZeblRXjfBGPCxzvsM/qRrfhO4cVWcZjhGJAH+ofzWT7Kmsxlu/GtE9hTG+lLWjps/3cmIJdBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f38194c49358bd42f15f8fe31545b07d848893f416046fc463029a7679cc3dc4","last_reissued_at":"2026-06-10T01:10:10.858573Z","signature_status":"signed_v1","first_computed_at":"2026-06-10T01:10:10.858573Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SARM2: Multi-Task Stage Aware Reward Modeling for Self Improving Robotic Manipulation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Chuan Wen, Hau Zheng, Jiankai Sun, Justin Yu, Ken Goldberg, Mac Schwager, Philipp Wu, Pieter Abbeel, Qianzhong Chen, Suning Huang, Yide Shentu","submitted_at":"2026-06-09T01:46:23Z","abstract_excerpt":"Fine-tuning vision-language-action (VLA) policies for long-horizon manipulation still relies heavily on behavior cloning, which requires costly high-quality demonstrations and keeps policies near the demonstration distribution. Reward models can reduce this dependence by reweighting demonstrations and providing dense supervision for on-robot reinforcement learning (RL), but they must be dense, accurate, and general. Existing methods fall short: task-specific stage-aware models are accurate but require per-task annotations, while general vision-language-model (VLM) reward models are broadly app"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.10305","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/2606.10305/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":"2606.10305","created_at":"2026-06-10T01:10:10.858723+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.10305v1","created_at":"2026-06-10T01:10:10.858723+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.10305","created_at":"2026-06-10T01:10:10.858723+00:00"},{"alias_kind":"pith_short_12","alias_value":"6OAZJRETLC6U","created_at":"2026-06-10T01:10:10.858723+00:00"},{"alias_kind":"pith_short_16","alias_value":"6OAZJRETLC6UF4K7","created_at":"2026-06-10T01:10:10.858723+00:00"},{"alias_kind":"pith_short_8","alias_value":"6OAZJRET","created_at":"2026-06-10T01:10:10.858723+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/6OAZJRETLC6UF4K7R7RRKRNQPW","json":"https://pith.science/pith/6OAZJRETLC6UF4K7R7RRKRNQPW.json","graph_json":"https://pith.science/api/pith-number/6OAZJRETLC6UF4K7R7RRKRNQPW/graph.json","events_json":"https://pith.science/api/pith-number/6OAZJRETLC6UF4K7R7RRKRNQPW/events.json","paper":"https://pith.science/paper/6OAZJRET"},"agent_actions":{"view_html":"https://pith.science/pith/6OAZJRETLC6UF4K7R7RRKRNQPW","download_json":"https://pith.science/pith/6OAZJRETLC6UF4K7R7RRKRNQPW.json","view_paper":"https://pith.science/paper/6OAZJRET","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.10305&json=true","fetch_graph":"https://pith.science/api/pith-number/6OAZJRETLC6UF4K7R7RRKRNQPW/graph.json","fetch_events":"https://pith.science/api/pith-number/6OAZJRETLC6UF4K7R7RRKRNQPW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6OAZJRETLC6UF4K7R7RRKRNQPW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6OAZJRETLC6UF4K7R7RRKRNQPW/action/storage_attestation","attest_author":"https://pith.science/pith/6OAZJRETLC6UF4K7R7RRKRNQPW/action/author_attestation","sign_citation":"https://pith.science/pith/6OAZJRETLC6UF4K7R7RRKRNQPW/action/citation_signature","submit_replication":"https://pith.science/pith/6OAZJRETLC6UF4K7R7RRKRNQPW/action/replication_record"}},"created_at":"2026-06-10T01:10:10.858723+00:00","updated_at":"2026-06-10T01:10:10.858723+00:00"}