{"paper":{"title":"parallelcbf: A composable safety-filter and auditability framework for tensor-parallel reinforcement learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"ParallelCBF unifies tensor-parallel UAV environments, hard-gate CBF safety filters, sharded BC-to-RL pipelines, and first-class operational auditability as composable APIs.","cross_cats":["cs.RO"],"primary_cat":"cs.LG","authors_text":"Yijun Lu, Yuyin Ma, Zilei Yang","submitted_at":"2026-05-15T01:06:16Z","abstract_excerpt":"While Isaac Lab provides massive parallel UAV simulation, OmniSafe and safe-control-gym provide constrained-RL benchmarks, and CBFKit provides control-barrier-function synthesis tooling, no existing framework unifies these capabilities for end-to-end safety-constrained training. ParallelCBF is the first framework to unify (i)~tensor-parallel UAV environments, (ii)~hard-gate CBF safety filters, (iii)~sharded BC-to-RL pipelines, and (iv)~first-class operational auditability -- pre-registration, watchdog registries, failure forensics, and dataset audits as composable APIs rather than user-impleme"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"ParallelCBF is the first framework to unify (i) tensor-parallel UAV environments, (ii) hard-gate CBF safety filters, (iii) sharded BC-to-RL pipelines, and (iv) first-class operational auditability as composable APIs rather than user-implemented scripts.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The premise that no prior framework already provides this specific unification of tensor-parallel simulation, hard CBF gating, BC-to-RL sharding, and integrated auditability primitives, and that embedding auditability as an architectural necessity (rather than optional script) is required for reproducible robotics research.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ParallelCBF is a composable framework that unifies tensor-parallel UAV environments, hard-gate CBF safety filters, sharded BC-to-RL pipelines, and operational auditability as first-class APIs for safe reinforcement learning.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ParallelCBF unifies tensor-parallel UAV environments, hard-gate CBF safety filters, sharded BC-to-RL pipelines, and first-class operational auditability as composable APIs.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"05ff9f059b5fe30a928e333e6409584dc9eb26e9449d9781c6ca32ccecd9c78e"},"source":{"id":"2605.15509","kind":"arxiv","version":1},"verdict":{"id":"3c0c5c6e-fc0d-438c-8015-ccd6f3d8673d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T15:47:11.070441Z","strongest_claim":"ParallelCBF is the first framework to unify (i) tensor-parallel UAV environments, (ii) hard-gate CBF safety filters, (iii) sharded BC-to-RL pipelines, and (iv) first-class operational auditability as composable APIs rather than user-implemented scripts.","one_line_summary":"ParallelCBF is a composable framework that unifies tensor-parallel UAV environments, hard-gate CBF safety filters, sharded BC-to-RL pipelines, and operational auditability as first-class APIs for safe reinforcement learning.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The premise that no prior framework already provides this specific unification of tensor-parallel simulation, hard CBF gating, BC-to-RL sharding, and integrated auditability primitives, and that embedding auditability as an architectural necessity (rather than optional script) is required for reproducible robotics research.","pith_extraction_headline":"ParallelCBF unifies tensor-parallel UAV environments, hard-gate CBF safety filters, sharded BC-to-RL pipelines, and first-class operational auditability as composable APIs."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15509/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T16:01:17.936100Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T15:53:35.223515Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T14:51:55.267702Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T14:21:54.057688Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"shingle_duplication","ran_at":"2026-05-19T13:49:41.850139Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"citation_quote_validity","ran_at":"2026-05-19T13:49:41.389010Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.636683Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"5ecba0d9bbf603cb66a3b7fd0ccaad514f75d8e831eab39eed56a0a2705c3e1d"},"references":{"count":14,"sample":[{"doi":"","year":2019,"title":"Control barrier functions: Theory and applications","work_id":"27c898ee-629a-4aa0-a512-c58fd9417a25","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Cbfkit: A control barrier function toolbox for robotics applications, 2024","work_id":"ecc665a0-dc62-4053-b6d3-1babed2c1942","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Mamba: Linear-time sequence modeling with selective state spaces, 2023","work_id":"e119cede-064a-4b64-9428-adf39f4521e3","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2018,"title":"Deep reinforcement learning that matters","work_id":"401903ab-1c03-4f44-9534-fad49cae4b57","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Safety gymnasium: A unified safe reinforcement learning benchmark","work_id":"7ca6b4a2-f185-4e5b-82b1-c176378a1c2c","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":14,"snapshot_sha256":"afd83e7d49939baabca8f871de30fa80a7c5c72702114fec86b486240b66c7d6","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"3fed2aa5d6e27b6d975e1924939d9f0e878775de8900a327f349714a14b39b1e"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}