{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:IZQRU7KHFEH2CL243OUAWULUR4","short_pith_number":"pith:IZQRU7KH","schema_version":"1.0","canonical_sha256":"46611a7d47290fa12f5cdba80b51748f02f3c4d65ef688c055c8cd2b4e1095f2","source":{"kind":"arxiv","id":"2606.04101","version":1},"attestation_state":"computed","paper":{"title":"UltraEP: Unleash MoE Training and Inference on Rack-Scale Nodes with Near-Optimal Load Balancing","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Bingyang Wu, Chao Jin, Chengxu Yang, Guojie Luo, Jing Mai, Qianchao Zhu, Shan Yu, Tuo Dai, Xinming Wei, Yinmin Zhong, Yuliang Liu, Zhouyang Li, Zili Zhang","submitted_at":"2026-06-02T18:07:51Z","abstract_excerpt":"Large-scale expert parallelism (EP) is becoming pivotal for training and serving frontier MoE models, but it also amplifies device-level expert load imbalance into compute stragglers, token all-to-all bottlenecks, and activation-memory spikes. Existing balancers redistribute experts periodically based on historical load, which becomes unreliable for production deployments with non-stationary load patterns.\n  We present UltraEP, the first exact-load, real-time balancer for large-EP MoE training and serving prefill on rack-scale nodes (RSNs). Built upon the extended scale-up connectivity of RSNs"},"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":"2606.04101","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2026-06-02T18:07:51Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a574e0ef8aaa8ac2164da80e964d012a8b7811194bc3ad85a23a656c817aec97","abstract_canon_sha256":"09f7980be28343aaef0adbc0cc8da8f0c53faa271dce8130f01c23ed2fac1209"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-04T00:06:46.723615Z","signature_b64":"GjsAmxmWllvzmmJxkl7jRrqEsKry7TUQv9mkcDY2eNCc4YoLyukwk4ZI2sI7edh65zBF+u4DNopuScdmbXy1Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"46611a7d47290fa12f5cdba80b51748f02f3c4d65ef688c055c8cd2b4e1095f2","last_reissued_at":"2026-06-04T00:06:46.723227Z","signature_status":"signed_v1","first_computed_at":"2026-06-04T00:06:46.723227Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"UltraEP: Unleash MoE Training and Inference on Rack-Scale Nodes with Near-Optimal Load Balancing","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Bingyang Wu, Chao Jin, Chengxu Yang, Guojie Luo, Jing Mai, Qianchao Zhu, Shan Yu, Tuo Dai, Xinming Wei, Yinmin Zhong, Yuliang Liu, Zhouyang Li, Zili Zhang","submitted_at":"2026-06-02T18:07:51Z","abstract_excerpt":"Large-scale expert parallelism (EP) is becoming pivotal for training and serving frontier MoE models, but it also amplifies device-level expert load imbalance into compute stragglers, token all-to-all bottlenecks, and activation-memory spikes. Existing balancers redistribute experts periodically based on historical load, which becomes unreliable for production deployments with non-stationary load patterns.\n  We present UltraEP, the first exact-load, real-time balancer for large-EP MoE training and serving prefill on rack-scale nodes (RSNs). Built upon the extended scale-up connectivity of RSNs"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.04101","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/2606.04101/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":"2606.04101","created_at":"2026-06-04T00:06:46.723288+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.04101v1","created_at":"2026-06-04T00:06:46.723288+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.04101","created_at":"2026-06-04T00:06:46.723288+00:00"},{"alias_kind":"pith_short_12","alias_value":"IZQRU7KHFEH2","created_at":"2026-06-04T00:06:46.723288+00:00"},{"alias_kind":"pith_short_16","alias_value":"IZQRU7KHFEH2CL24","created_at":"2026-06-04T00:06:46.723288+00:00"},{"alias_kind":"pith_short_8","alias_value":"IZQRU7KH","created_at":"2026-06-04T00:06:46.723288+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/IZQRU7KHFEH2CL243OUAWULUR4","json":"https://pith.science/pith/IZQRU7KHFEH2CL243OUAWULUR4.json","graph_json":"https://pith.science/api/pith-number/IZQRU7KHFEH2CL243OUAWULUR4/graph.json","events_json":"https://pith.science/api/pith-number/IZQRU7KHFEH2CL243OUAWULUR4/events.json","paper":"https://pith.science/paper/IZQRU7KH"},"agent_actions":{"view_html":"https://pith.science/pith/IZQRU7KHFEH2CL243OUAWULUR4","download_json":"https://pith.science/pith/IZQRU7KHFEH2CL243OUAWULUR4.json","view_paper":"https://pith.science/paper/IZQRU7KH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.04101&json=true","fetch_graph":"https://pith.science/api/pith-number/IZQRU7KHFEH2CL243OUAWULUR4/graph.json","fetch_events":"https://pith.science/api/pith-number/IZQRU7KHFEH2CL243OUAWULUR4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IZQRU7KHFEH2CL243OUAWULUR4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IZQRU7KHFEH2CL243OUAWULUR4/action/storage_attestation","attest_author":"https://pith.science/pith/IZQRU7KHFEH2CL243OUAWULUR4/action/author_attestation","sign_citation":"https://pith.science/pith/IZQRU7KHFEH2CL243OUAWULUR4/action/citation_signature","submit_replication":"https://pith.science/pith/IZQRU7KHFEH2CL243OUAWULUR4/action/replication_record"}},"created_at":"2026-06-04T00:06:46.723288+00:00","updated_at":"2026-06-04T00:06:46.723288+00:00"}