{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WSOUZLYLQCDIDOJJYZEFEQLXEX","short_pith_number":"pith:WSOUZLYL","schema_version":"1.0","canonical_sha256":"b49d4caf0b808681b929c64852417725c496c9269ba9cefe0da9582805f72deb","source":{"kind":"arxiv","id":"2505.05950","version":2},"attestation_state":"computed","paper":{"title":"FloE: On-the-Fly MoE Inference on Memory-constrained GPU","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Jue Wang, Jun Zhang, Ke Chen, Lidan Shou, Yiping Wang, Yuxin Zhou, Zheng Li, Zhongle Xie","submitted_at":"2025-05-09T10:53:47Z","abstract_excerpt":"With the widespread adoption of Mixture-of-Experts (MoE) models, there is a growing demand for efficient inference on memory-constrained devices. While offloading expert parameters to CPU memory and loading activated experts on demand has emerged as a potential solution, the large size of activated experts overburdens the limited PCIe bandwidth, hindering the effectiveness in latency-sensitive scenarios. To mitigate this, we propose FloE, an on-the-fly MoE inference system on memory-constrained GPUs. FloE is built on the insight that there exists substantial untapped redundancy within sparsely"},"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":"2505.05950","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-09T10:53:47Z","cross_cats_sorted":[],"title_canon_sha256":"0cce9fca24973be8a517078b37782cac914db0d3c85179879ddfcf0abcdb55cb","abstract_canon_sha256":"d4d9bef36b9134d7310cba5971c99f41a118071b27018207e8b3c1091ad8ea4e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:01:37.221383Z","signature_b64":"bQliAxe087cdFR/zCzPoFRqhxlcToHvpzd/Y/rNblOF6FzY20WBqmFnZrHRrVMThBCzBIrrpBVWf379XBM3rDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b49d4caf0b808681b929c64852417725c496c9269ba9cefe0da9582805f72deb","last_reissued_at":"2026-07-05T11:01:37.220895Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:01:37.220895Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FloE: On-the-Fly MoE Inference on Memory-constrained GPU","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Jue Wang, Jun Zhang, Ke Chen, Lidan Shou, Yiping Wang, Yuxin Zhou, Zheng Li, Zhongle Xie","submitted_at":"2025-05-09T10:53:47Z","abstract_excerpt":"With the widespread adoption of Mixture-of-Experts (MoE) models, there is a growing demand for efficient inference on memory-constrained devices. While offloading expert parameters to CPU memory and loading activated experts on demand has emerged as a potential solution, the large size of activated experts overburdens the limited PCIe bandwidth, hindering the effectiveness in latency-sensitive scenarios. To mitigate this, we propose FloE, an on-the-fly MoE inference system on memory-constrained GPUs. FloE is built on the insight that there exists substantial untapped redundancy within sparsely"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.05950","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/2505.05950/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":"2505.05950","created_at":"2026-07-05T11:01:37.220962+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.05950v2","created_at":"2026-07-05T11:01:37.220962+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.05950","created_at":"2026-07-05T11:01:37.220962+00:00"},{"alias_kind":"pith_short_12","alias_value":"WSOUZLYLQCDI","created_at":"2026-07-05T11:01:37.220962+00:00"},{"alias_kind":"pith_short_16","alias_value":"WSOUZLYLQCDIDOJJ","created_at":"2026-07-05T11:01:37.220962+00:00"},{"alias_kind":"pith_short_8","alias_value":"WSOUZLYL","created_at":"2026-07-05T11:01:37.220962+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.26881","citing_title":"FaaSMoE: A Serverless Framework for Multi-Tenant Mixture-of-Experts Serving","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WSOUZLYLQCDIDOJJYZEFEQLXEX","json":"https://pith.science/pith/WSOUZLYLQCDIDOJJYZEFEQLXEX.json","graph_json":"https://pith.science/api/pith-number/WSOUZLYLQCDIDOJJYZEFEQLXEX/graph.json","events_json":"https://pith.science/api/pith-number/WSOUZLYLQCDIDOJJYZEFEQLXEX/events.json","paper":"https://pith.science/paper/WSOUZLYL"},"agent_actions":{"view_html":"https://pith.science/pith/WSOUZLYLQCDIDOJJYZEFEQLXEX","download_json":"https://pith.science/pith/WSOUZLYLQCDIDOJJYZEFEQLXEX.json","view_paper":"https://pith.science/paper/WSOUZLYL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.05950&json=true","fetch_graph":"https://pith.science/api/pith-number/WSOUZLYLQCDIDOJJYZEFEQLXEX/graph.json","fetch_events":"https://pith.science/api/pith-number/WSOUZLYLQCDIDOJJYZEFEQLXEX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WSOUZLYLQCDIDOJJYZEFEQLXEX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WSOUZLYLQCDIDOJJYZEFEQLXEX/action/storage_attestation","attest_author":"https://pith.science/pith/WSOUZLYLQCDIDOJJYZEFEQLXEX/action/author_attestation","sign_citation":"https://pith.science/pith/WSOUZLYLQCDIDOJJYZEFEQLXEX/action/citation_signature","submit_replication":"https://pith.science/pith/WSOUZLYLQCDIDOJJYZEFEQLXEX/action/replication_record"}},"created_at":"2026-07-05T11:01:37.220962+00:00","updated_at":"2026-07-05T11:01:37.220962+00:00"}