{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:A7VYGNGVTJ4XOOZFSKE56Q2N35","short_pith_number":"pith:A7VYGNGV","schema_version":"1.0","canonical_sha256":"07eb8334d59a79773b259289df434ddf62194727bc1f07e82c9268220a9b7390","source":{"kind":"arxiv","id":"2607.18722","version":1},"attestation_state":"computed","paper":{"title":"Stale but Stable: Staleness-Adaptive Trust Regions for Stabilizing Asynchronous Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Haitao Mi, Junyao Yang, Kishan Panaganti, Leowei Liang, Ruhan Wang, Xiangxin Zhou, Yucheng Shi, Zhongzhi Li, Zongxia Li","submitted_at":"2026-07-21T05:27:50Z","abstract_excerpt":"Asynchronous reinforcement learning improves throughput by decoupling rollout generation from optimization, but staleness is an inevitable byproduct compounded by policy lag, engine delays, and mixture-of-experts routing. From a trust-region perspective, this mismatch is critical: training-inference divergence governs approximation error in finite-horizon bounds, whereas PPO clipping only gates sampled outward updates, acting as a sampled surrogate rather than a full-policy constraint. As a result, high-staleness updates remain weakly controlled in the asynchronous regime where stale rollouts "},"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":"2607.18722","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-07-21T05:27:50Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"5596c7a82da4d28eda75ca80f44114b88bcdd4b12624a37f9dc8546e4a3cb3ee","abstract_canon_sha256":"8a9762f9ef8ea3e4d08f304d5293d827ba9cf0d7284dc6eaceffc6d72eb94744"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-22T01:23:08.486598Z","signature_b64":"K41WG290Dztm1hp613tLLHVc3x1G++OdIf4YqZ8vvgDvorzFUboFM1k79AuD+OMFRV6YbwBVBRleRi0Js9qyAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"07eb8334d59a79773b259289df434ddf62194727bc1f07e82c9268220a9b7390","last_reissued_at":"2026-07-22T01:23:08.485702Z","signature_status":"signed_v1","first_computed_at":"2026-07-22T01:23:08.485702Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Stale but Stable: Staleness-Adaptive Trust Regions for Stabilizing Asynchronous Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Haitao Mi, Junyao Yang, Kishan Panaganti, Leowei Liang, Ruhan Wang, Xiangxin Zhou, Yucheng Shi, Zhongzhi Li, Zongxia Li","submitted_at":"2026-07-21T05:27:50Z","abstract_excerpt":"Asynchronous reinforcement learning improves throughput by decoupling rollout generation from optimization, but staleness is an inevitable byproduct compounded by policy lag, engine delays, and mixture-of-experts routing. From a trust-region perspective, this mismatch is critical: training-inference divergence governs approximation error in finite-horizon bounds, whereas PPO clipping only gates sampled outward updates, acting as a sampled surrogate rather than a full-policy constraint. As a result, high-staleness updates remain weakly controlled in the asynchronous regime where stale rollouts "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.18722","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/2607.18722/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":"2607.18722","created_at":"2026-07-22T01:23:08.486157+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.18722v1","created_at":"2026-07-22T01:23:08.486157+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.18722","created_at":"2026-07-22T01:23:08.486157+00:00"},{"alias_kind":"pith_short_12","alias_value":"A7VYGNGVTJ4X","created_at":"2026-07-22T01:23:08.486157+00:00"},{"alias_kind":"pith_short_16","alias_value":"A7VYGNGVTJ4XOOZF","created_at":"2026-07-22T01:23:08.486157+00:00"},{"alias_kind":"pith_short_8","alias_value":"A7VYGNGV","created_at":"2026-07-22T01:23:08.486157+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/A7VYGNGVTJ4XOOZFSKE56Q2N35","json":"https://pith.science/pith/A7VYGNGVTJ4XOOZFSKE56Q2N35.json","graph_json":"https://pith.science/api/pith-number/A7VYGNGVTJ4XOOZFSKE56Q2N35/graph.json","events_json":"https://pith.science/api/pith-number/A7VYGNGVTJ4XOOZFSKE56Q2N35/events.json","paper":"https://pith.science/paper/A7VYGNGV"},"agent_actions":{"view_html":"https://pith.science/pith/A7VYGNGVTJ4XOOZFSKE56Q2N35","download_json":"https://pith.science/pith/A7VYGNGVTJ4XOOZFSKE56Q2N35.json","view_paper":"https://pith.science/paper/A7VYGNGV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.18722&json=true","fetch_graph":"https://pith.science/api/pith-number/A7VYGNGVTJ4XOOZFSKE56Q2N35/graph.json","fetch_events":"https://pith.science/api/pith-number/A7VYGNGVTJ4XOOZFSKE56Q2N35/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/A7VYGNGVTJ4XOOZFSKE56Q2N35/action/timestamp_anchor","attest_storage":"https://pith.science/pith/A7VYGNGVTJ4XOOZFSKE56Q2N35/action/storage_attestation","attest_author":"https://pith.science/pith/A7VYGNGVTJ4XOOZFSKE56Q2N35/action/author_attestation","sign_citation":"https://pith.science/pith/A7VYGNGVTJ4XOOZFSKE56Q2N35/action/citation_signature","submit_replication":"https://pith.science/pith/A7VYGNGVTJ4XOOZFSKE56Q2N35/action/replication_record"}},"created_at":"2026-07-22T01:23:08.486157+00:00","updated_at":"2026-07-22T01:23:08.486157+00:00"}