{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:6V524WSHOGJTNFWMKDTUTMMHDS","short_pith_number":"pith:6V524WSH","schema_version":"1.0","canonical_sha256":"f57bae5a4771933696cc50e749b1871c9eba86d44b0ce585a2819a22fe114bef","source":{"kind":"arxiv","id":"2606.11867","version":1},"attestation_state":"computed","paper":{"title":"Harnessing Routing Foresight for Micro-step-level MoE load balancing in RL Post-training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Bin Cui, Fangcheng Fu, Haoyang Li, Jie Jiang, Sheng Lin, Tong Zhao, Xupeng Miao, Yanfeng Zhao, Yuming Zhou","submitted_at":"2026-06-10T09:42:11Z","abstract_excerpt":"Mixture-of-Experts (MoE) and reinforcement learning (RL) post-training now dominate large language model (LLM) development, yet expert load imbalance remains a critical challenge. Existing load-balancing systems target pre-training by relying on historical step-level statistics. However, these methods fail under the unique workload dynamics of RL post-training: the step-level load is stable, but the tiny batch sizes processed during micro-steps cause severe, high-frequency load fluctuations.\n  We introduce ForeMoE, a micro-step-level load balancing system for MoE RL post-training. Instead of r"},"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":"2606.11867","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2026-06-10T09:42:11Z","cross_cats_sorted":[],"title_canon_sha256":"5423552128a94fe208aab6d89d93db5caf7e4c610f6c752bfb13dcaddb01531a","abstract_canon_sha256":"f4e5ad8da76d9ea8465b76130e80a7004433a67b1e71b81856200da5f09894ef"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-11T01:10:12.667593Z","signature_b64":"H9EQB5qN35lkzKBLXFG9G7micMX6gf9bv9UlGFwYh/rylRToy7QMLdu3T2DRvSepniYJwXqEaTTwQK53M3k+Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f57bae5a4771933696cc50e749b1871c9eba86d44b0ce585a2819a22fe114bef","last_reissued_at":"2026-06-11T01:10:12.666808Z","signature_status":"signed_v1","first_computed_at":"2026-06-11T01:10:12.666808Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Harnessing Routing Foresight for Micro-step-level MoE load balancing in RL Post-training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Bin Cui, Fangcheng Fu, Haoyang Li, Jie Jiang, Sheng Lin, Tong Zhao, Xupeng Miao, Yanfeng Zhao, Yuming Zhou","submitted_at":"2026-06-10T09:42:11Z","abstract_excerpt":"Mixture-of-Experts (MoE) and reinforcement learning (RL) post-training now dominate large language model (LLM) development, yet expert load imbalance remains a critical challenge. Existing load-balancing systems target pre-training by relying on historical step-level statistics. However, these methods fail under the unique workload dynamics of RL post-training: the step-level load is stable, but the tiny batch sizes processed during micro-steps cause severe, high-frequency load fluctuations.\n  We introduce ForeMoE, a micro-step-level load balancing system for MoE RL post-training. Instead of r"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.11867","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/2606.11867/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":"2606.11867","created_at":"2026-06-11T01:10:12.666934+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.11867v1","created_at":"2026-06-11T01:10:12.666934+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.11867","created_at":"2026-06-11T01:10:12.666934+00:00"},{"alias_kind":"pith_short_12","alias_value":"6V524WSHOGJT","created_at":"2026-06-11T01:10:12.666934+00:00"},{"alias_kind":"pith_short_16","alias_value":"6V524WSHOGJTNFWM","created_at":"2026-06-11T01:10:12.666934+00:00"},{"alias_kind":"pith_short_8","alias_value":"6V524WSH","created_at":"2026-06-11T01:10:12.666934+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/6V524WSHOGJTNFWMKDTUTMMHDS","json":"https://pith.science/pith/6V524WSHOGJTNFWMKDTUTMMHDS.json","graph_json":"https://pith.science/api/pith-number/6V524WSHOGJTNFWMKDTUTMMHDS/graph.json","events_json":"https://pith.science/api/pith-number/6V524WSHOGJTNFWMKDTUTMMHDS/events.json","paper":"https://pith.science/paper/6V524WSH"},"agent_actions":{"view_html":"https://pith.science/pith/6V524WSHOGJTNFWMKDTUTMMHDS","download_json":"https://pith.science/pith/6V524WSHOGJTNFWMKDTUTMMHDS.json","view_paper":"https://pith.science/paper/6V524WSH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.11867&json=true","fetch_graph":"https://pith.science/api/pith-number/6V524WSHOGJTNFWMKDTUTMMHDS/graph.json","fetch_events":"https://pith.science/api/pith-number/6V524WSHOGJTNFWMKDTUTMMHDS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6V524WSHOGJTNFWMKDTUTMMHDS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6V524WSHOGJTNFWMKDTUTMMHDS/action/storage_attestation","attest_author":"https://pith.science/pith/6V524WSHOGJTNFWMKDTUTMMHDS/action/author_attestation","sign_citation":"https://pith.science/pith/6V524WSHOGJTNFWMKDTUTMMHDS/action/citation_signature","submit_replication":"https://pith.science/pith/6V524WSHOGJTNFWMKDTUTMMHDS/action/replication_record"}},"created_at":"2026-06-11T01:10:12.666934+00:00","updated_at":"2026-06-11T01:10:12.666934+00:00"}