{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:4KOVMX3JEOUWMWATOXKQGVJ55Z","short_pith_number":"pith:4KOVMX3J","schema_version":"1.0","canonical_sha256":"e29d565f6923a966581375d503553dee500c6afcf01c8e5532b3b10c4df4b0c7","source":{"kind":"arxiv","id":"2606.03070","version":1},"attestation_state":"computed","paper":{"title":"ASymPO: Asymmetric-Scale Policy Optimization for Asynchronous LLM Post-Training Without Behavior Information","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Mingxuan Yuan, Tao Zhong, Xiaojin Fu, Yuxuan Yao, Zehua Liu","submitted_at":"2026-06-02T03:00:34Z","abstract_excerpt":"Asynchronous reinforcement learning can improve language-model post-training throughput by decoupling response generation from policy optimization, but stale responses introduce distribution drift. Standard behavior-corrected methods control this drift with behavior-policy probabilities, importance ratios, or clipping, which requires token-aligned, versioned, and numerically consistent behavior log-probabilities across rollout and learner systems. We ask whether asynchronous group-relative RL can instead be stabilized using only current-policy probabilities. We identify a scale-imbalance failu"},"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":"2606.03070","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-02T03:00:34Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a370bf5dddec016818490cb3694cca5a19e27ace727b434995201e53370a6683","abstract_canon_sha256":"ae18d5cae15780264c9fedd4922ab592b899ca04e4d3f841dc7ed565a2f04aee"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-03T01:05:30.859248Z","signature_b64":"WVEX7vyW/SmlINfVWzb33u6izxui9P2eei58WXScrBLbf4m56u9OThhNCK+GaEMiQpiq/xxcFpkRyEzsvfyRCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e29d565f6923a966581375d503553dee500c6afcf01c8e5532b3b10c4df4b0c7","last_reissued_at":"2026-06-03T01:05:30.858790Z","signature_status":"signed_v1","first_computed_at":"2026-06-03T01:05:30.858790Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ASymPO: Asymmetric-Scale Policy Optimization for Asynchronous LLM Post-Training Without Behavior Information","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Mingxuan Yuan, Tao Zhong, Xiaojin Fu, Yuxuan Yao, Zehua Liu","submitted_at":"2026-06-02T03:00:34Z","abstract_excerpt":"Asynchronous reinforcement learning can improve language-model post-training throughput by decoupling response generation from policy optimization, but stale responses introduce distribution drift. Standard behavior-corrected methods control this drift with behavior-policy probabilities, importance ratios, or clipping, which requires token-aligned, versioned, and numerically consistent behavior log-probabilities across rollout and learner systems. We ask whether asynchronous group-relative RL can instead be stabilized using only current-policy probabilities. We identify a scale-imbalance failu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.03070","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/2606.03070/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":"2606.03070","created_at":"2026-06-03T01:05:30.858870+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.03070v1","created_at":"2026-06-03T01:05:30.858870+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.03070","created_at":"2026-06-03T01:05:30.858870+00:00"},{"alias_kind":"pith_short_12","alias_value":"4KOVMX3JEOUW","created_at":"2026-06-03T01:05:30.858870+00:00"},{"alias_kind":"pith_short_16","alias_value":"4KOVMX3JEOUWMWAT","created_at":"2026-06-03T01:05:30.858870+00:00"},{"alias_kind":"pith_short_8","alias_value":"4KOVMX3J","created_at":"2026-06-03T01:05:30.858870+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/4KOVMX3JEOUWMWATOXKQGVJ55Z","json":"https://pith.science/pith/4KOVMX3JEOUWMWATOXKQGVJ55Z.json","graph_json":"https://pith.science/api/pith-number/4KOVMX3JEOUWMWATOXKQGVJ55Z/graph.json","events_json":"https://pith.science/api/pith-number/4KOVMX3JEOUWMWATOXKQGVJ55Z/events.json","paper":"https://pith.science/paper/4KOVMX3J"},"agent_actions":{"view_html":"https://pith.science/pith/4KOVMX3JEOUWMWATOXKQGVJ55Z","download_json":"https://pith.science/pith/4KOVMX3JEOUWMWATOXKQGVJ55Z.json","view_paper":"https://pith.science/paper/4KOVMX3J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.03070&json=true","fetch_graph":"https://pith.science/api/pith-number/4KOVMX3JEOUWMWATOXKQGVJ55Z/graph.json","fetch_events":"https://pith.science/api/pith-number/4KOVMX3JEOUWMWATOXKQGVJ55Z/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4KOVMX3JEOUWMWATOXKQGVJ55Z/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4KOVMX3JEOUWMWATOXKQGVJ55Z/action/storage_attestation","attest_author":"https://pith.science/pith/4KOVMX3JEOUWMWATOXKQGVJ55Z/action/author_attestation","sign_citation":"https://pith.science/pith/4KOVMX3JEOUWMWATOXKQGVJ55Z/action/citation_signature","submit_replication":"https://pith.science/pith/4KOVMX3JEOUWMWATOXKQGVJ55Z/action/replication_record"}},"created_at":"2026-06-03T01:05:30.858870+00:00","updated_at":"2026-06-03T01:05:30.858870+00:00"}