{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:E3GKMRTWMMAUFLQVT7L2PNKEO3","short_pith_number":"pith:E3GKMRTW","schema_version":"1.0","canonical_sha256":"26cca64676630142ae159fd7a7b54476e49c6540e676b03ed3de8a53945b999f","source":{"kind":"arxiv","id":"2608.13026","version":1},"attestation_state":"computed","paper":{"title":"Temporal GRPO: Beyond Trajectory-Level Credit in Vision-Language-Action Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Changwen Zheng, Fengge Wu, Hang Gao, Wenwen Qiang, Yao Zhou","submitted_at":"2026-08-13T09:54:14Z","abstract_excerpt":"Outcome-driven reinforcement learning offers a scalable way to post-train vision-language-action (VLA) policies from sparse task-success feedback. In common GRPO-based VLA post-training, one rollout-level advantage is applied to every action in the trajectory. A rollout that completes several valid stages but fails later can therefore penalize the actions that produced its earlier progress. We call this trajectory-level credit aliasing. Temporal GRPO addresses this problem by constructing detectable task stages, aligning each rollout with stage-specific action intervals, and comparing only rol"},"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":"2608.13026","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-08-13T09:54:14Z","cross_cats_sorted":[],"title_canon_sha256":"a77512911f3617c30cb81b6060ba4b14af0574f21c3766b39c8dfdceb13dba25","abstract_canon_sha256":"bb294a27ee73cbfcb3efdd229d8f02325e1a3577aa8c4e181c6e6256ed182d66"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-14T01:00:09.326859Z","signature_b64":"+ECAVUFBNrL7whgm2sqW/IxycpZTELCJrP1Xk4odp28ORtmbtmNRdeVafVHbPuopsFoat2DyO0KQ+eVnWwLnBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"26cca64676630142ae159fd7a7b54476e49c6540e676b03ed3de8a53945b999f","last_reissued_at":"2026-08-14T01:00:09.324829Z","signature_status":"signed_v1","first_computed_at":"2026-08-14T01:00:09.324829Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Temporal GRPO: Beyond Trajectory-Level Credit in Vision-Language-Action Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Changwen Zheng, Fengge Wu, Hang Gao, Wenwen Qiang, Yao Zhou","submitted_at":"2026-08-13T09:54:14Z","abstract_excerpt":"Outcome-driven reinforcement learning offers a scalable way to post-train vision-language-action (VLA) policies from sparse task-success feedback. In common GRPO-based VLA post-training, one rollout-level advantage is applied to every action in the trajectory. A rollout that completes several valid stages but fails later can therefore penalize the actions that produced its earlier progress. We call this trajectory-level credit aliasing. Temporal GRPO addresses this problem by constructing detectable task stages, aligning each rollout with stage-specific action intervals, and comparing only rol"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.13026","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/2608.13026/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":"2608.13026","created_at":"2026-08-14T01:00:09.325830+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.13026v1","created_at":"2026-08-14T01:00:09.325830+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.13026","created_at":"2026-08-14T01:00:09.325830+00:00"},{"alias_kind":"pith_short_12","alias_value":"E3GKMRTWMMAU","created_at":"2026-08-14T01:00:09.325830+00:00"},{"alias_kind":"pith_short_16","alias_value":"E3GKMRTWMMAUFLQV","created_at":"2026-08-14T01:00:09.325830+00:00"},{"alias_kind":"pith_short_8","alias_value":"E3GKMRTW","created_at":"2026-08-14T01:00:09.325830+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/E3GKMRTWMMAUFLQVT7L2PNKEO3","json":"https://pith.science/pith/E3GKMRTWMMAUFLQVT7L2PNKEO3.json","graph_json":"https://pith.science/api/pith-number/E3GKMRTWMMAUFLQVT7L2PNKEO3/graph.json","events_json":"https://pith.science/api/pith-number/E3GKMRTWMMAUFLQVT7L2PNKEO3/events.json","paper":"https://pith.science/paper/E3GKMRTW"},"agent_actions":{"view_html":"https://pith.science/pith/E3GKMRTWMMAUFLQVT7L2PNKEO3","download_json":"https://pith.science/pith/E3GKMRTWMMAUFLQVT7L2PNKEO3.json","view_paper":"https://pith.science/paper/E3GKMRTW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.13026&json=true","fetch_graph":"https://pith.science/api/pith-number/E3GKMRTWMMAUFLQVT7L2PNKEO3/graph.json","fetch_events":"https://pith.science/api/pith-number/E3GKMRTWMMAUFLQVT7L2PNKEO3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/E3GKMRTWMMAUFLQVT7L2PNKEO3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/E3GKMRTWMMAUFLQVT7L2PNKEO3/action/storage_attestation","attest_author":"https://pith.science/pith/E3GKMRTWMMAUFLQVT7L2PNKEO3/action/author_attestation","sign_citation":"https://pith.science/pith/E3GKMRTWMMAUFLQVT7L2PNKEO3/action/citation_signature","submit_replication":"https://pith.science/pith/E3GKMRTWMMAUFLQVT7L2PNKEO3/action/replication_record"}},"created_at":"2026-08-14T01:00:09.325830+00:00","updated_at":"2026-08-14T01:00:09.325830+00:00"}