{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RDLM25SLJXHBRZGFYPHRVOD4XZ","short_pith_number":"pith:RDLM25SL","schema_version":"1.0","canonical_sha256":"88d6cd764b4dce18e4c5c3cf1ab87cbe4818df349e2cfdc563e0173f7d54c8f9","source":{"kind":"arxiv","id":"2603.14371","version":2},"attestation_state":"computed","paper":{"title":"OxyGen: Unified KV Cache Management for VLA Inference under Multi-Task Parallelism","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.RO","authors_text":"Huaizhi Tang, Ting Cao, Weijun Wang, Xiangyu Li, Xin Ding, Yunxin Liu","submitted_at":"2026-03-15T13:23:56Z","abstract_excerpt":"Embodied AI agents increasingly require parallel execution of multiple tasks, such as manipulation, conversation, and memory construction, from shared observations under distinct time constraints. Recent Mixture-of-Transformers (MoT) Vision-Language-Action Models (VLAs) architecturally support such heterogeneous outputs, yet existing inference systems fail to achieve efficient multi-task parallelism for on-device deployment because of redundant computation and resource contention. We identify isolated KV cache management as the root cause. To address this, we propose unified KV cache managemen"},"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":"2603.14371","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-03-15T13:23:56Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c256bbbc84fb2e41d8ef0ae892272bf9af49b158f65791fbcf14935f7127be5e","abstract_canon_sha256":"508c47fa130ac0bfe1daae495835c2333be3791ebe8d0d1db4f69984b2194120"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:05:43.325129Z","signature_b64":"5+eZ2cYYXE1B2dKSNw9fKJnbVHrYAb7yCYd8ONQckqFOPh6K9K+1BMx1/b/YQiQaCL0OAvGghPZhxa2D6iFZDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"88d6cd764b4dce18e4c5c3cf1ab87cbe4818df349e2cfdc563e0173f7d54c8f9","last_reissued_at":"2026-05-20T00:05:43.324610Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:05:43.324610Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"OxyGen: Unified KV Cache Management for VLA Inference under Multi-Task Parallelism","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.RO","authors_text":"Huaizhi Tang, Ting Cao, Weijun Wang, Xiangyu Li, Xin Ding, Yunxin Liu","submitted_at":"2026-03-15T13:23:56Z","abstract_excerpt":"Embodied AI agents increasingly require parallel execution of multiple tasks, such as manipulation, conversation, and memory construction, from shared observations under distinct time constraints. Recent Mixture-of-Transformers (MoT) Vision-Language-Action Models (VLAs) architecturally support such heterogeneous outputs, yet existing inference systems fail to achieve efficient multi-task parallelism for on-device deployment because of redundant computation and resource contention. We identify isolated KV cache management as the root cause. To address this, we propose unified KV cache managemen"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.14371","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/2603.14371/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":"2603.14371","created_at":"2026-05-20T00:05:43.324692+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.14371v2","created_at":"2026-05-20T00:05:43.324692+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.14371","created_at":"2026-05-20T00:05:43.324692+00:00"},{"alias_kind":"pith_short_12","alias_value":"RDLM25SLJXHB","created_at":"2026-05-20T00:05:43.324692+00:00"},{"alias_kind":"pith_short_16","alias_value":"RDLM25SLJXHBRZGF","created_at":"2026-05-20T00:05:43.324692+00:00"},{"alias_kind":"pith_short_8","alias_value":"RDLM25SL","created_at":"2026-05-20T00:05:43.324692+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.19294","citing_title":"DEFLECT: Delay-Robust Execution via Flow-matching Likelihood-Estimated Counterfactual Tuning for VLA Policies","ref_index":28,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RDLM25SLJXHBRZGFYPHRVOD4XZ","json":"https://pith.science/pith/RDLM25SLJXHBRZGFYPHRVOD4XZ.json","graph_json":"https://pith.science/api/pith-number/RDLM25SLJXHBRZGFYPHRVOD4XZ/graph.json","events_json":"https://pith.science/api/pith-number/RDLM25SLJXHBRZGFYPHRVOD4XZ/events.json","paper":"https://pith.science/paper/RDLM25SL"},"agent_actions":{"view_html":"https://pith.science/pith/RDLM25SLJXHBRZGFYPHRVOD4XZ","download_json":"https://pith.science/pith/RDLM25SLJXHBRZGFYPHRVOD4XZ.json","view_paper":"https://pith.science/paper/RDLM25SL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.14371&json=true","fetch_graph":"https://pith.science/api/pith-number/RDLM25SLJXHBRZGFYPHRVOD4XZ/graph.json","fetch_events":"https://pith.science/api/pith-number/RDLM25SLJXHBRZGFYPHRVOD4XZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RDLM25SLJXHBRZGFYPHRVOD4XZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RDLM25SLJXHBRZGFYPHRVOD4XZ/action/storage_attestation","attest_author":"https://pith.science/pith/RDLM25SLJXHBRZGFYPHRVOD4XZ/action/author_attestation","sign_citation":"https://pith.science/pith/RDLM25SLJXHBRZGFYPHRVOD4XZ/action/citation_signature","submit_replication":"https://pith.science/pith/RDLM25SLJXHBRZGFYPHRVOD4XZ/action/replication_record"}},"created_at":"2026-05-20T00:05:43.324692+00:00","updated_at":"2026-05-20T00:05:43.324692+00:00"}