{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:TPND2KRXWWBMNZWLSA4BQSGOUW","short_pith_number":"pith:TPND2KRX","schema_version":"1.0","canonical_sha256":"9bda3d2a37b582c6e6cb90381848cea5a9530b7eb36f0a53568e8ba09662f4b6","source":{"kind":"arxiv","id":"2504.09345","version":1},"attestation_state":"computed","paper":{"title":"MoE-Lens: Towards the Hardware Limit of High-Throughput MoE LLM Serving Under Resource Constraints","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.DC","authors_text":"Lin Ma, Nishil Talati, Yichao Yuan","submitted_at":"2025-04-12T21:26:56Z","abstract_excerpt":"Mixture of Experts (MoE) LLMs, characterized by their sparse activation patterns, offer a promising approach to scaling language models while avoiding proportionally increasing the inference cost. However, their large parameter sizes present deployment challenges in resource-constrained environments with limited GPU memory capacity, as GPU memory is often insufficient to accommodate the full set of model weights. Consequently, typical deployments rely on CPU-GPU hybrid execution: the GPU handles compute-intensive GEMM operations, while the CPU processes the relatively lightweight attention mec"},"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":"2504.09345","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2025-04-12T21:26:56Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"875e46653347ae96c3042e82782e4869f44e64edd48dc0cab821a18024759912","abstract_canon_sha256":"7ab9bb5fc954833aa785acadcf4c5174efa374d51530ede7e8eb4c5523936d0a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:48:35.885660Z","signature_b64":"01C/gdQtHwfjWqVxh/yBOH0/emZxOO9StWQNsYJ18f2FxDXGQzcNi2xOqOjUyh2N+MTjs9+gkp1EYpbm4VM3Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9bda3d2a37b582c6e6cb90381848cea5a9530b7eb36f0a53568e8ba09662f4b6","last_reissued_at":"2026-07-05T10:48:35.885191Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:48:35.885191Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MoE-Lens: Towards the Hardware Limit of High-Throughput MoE LLM Serving Under Resource Constraints","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.DC","authors_text":"Lin Ma, Nishil Talati, Yichao Yuan","submitted_at":"2025-04-12T21:26:56Z","abstract_excerpt":"Mixture of Experts (MoE) LLMs, characterized by their sparse activation patterns, offer a promising approach to scaling language models while avoiding proportionally increasing the inference cost. However, their large parameter sizes present deployment challenges in resource-constrained environments with limited GPU memory capacity, as GPU memory is often insufficient to accommodate the full set of model weights. Consequently, typical deployments rely on CPU-GPU hybrid execution: the GPU handles compute-intensive GEMM operations, while the CPU processes the relatively lightweight attention mec"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.09345","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/2504.09345/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":"2504.09345","created_at":"2026-07-05T10:48:35.885250+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.09345v1","created_at":"2026-07-05T10:48:35.885250+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.09345","created_at":"2026-07-05T10:48:35.885250+00:00"},{"alias_kind":"pith_short_12","alias_value":"TPND2KRXWWBM","created_at":"2026-07-05T10:48:35.885250+00:00"},{"alias_kind":"pith_short_16","alias_value":"TPND2KRXWWBMNZWL","created_at":"2026-07-05T10:48:35.885250+00:00"},{"alias_kind":"pith_short_8","alias_value":"TPND2KRX","created_at":"2026-07-05T10:48:35.885250+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.14626","citing_title":"ELMoE-3D: Leveraging Intrinsic Elasticity of MoE for Hybrid-Bonding-Enabled Self-Speculative Decoding in On-Premises Serving","ref_index":73,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TPND2KRXWWBMNZWLSA4BQSGOUW","json":"https://pith.science/pith/TPND2KRXWWBMNZWLSA4BQSGOUW.json","graph_json":"https://pith.science/api/pith-number/TPND2KRXWWBMNZWLSA4BQSGOUW/graph.json","events_json":"https://pith.science/api/pith-number/TPND2KRXWWBMNZWLSA4BQSGOUW/events.json","paper":"https://pith.science/paper/TPND2KRX"},"agent_actions":{"view_html":"https://pith.science/pith/TPND2KRXWWBMNZWLSA4BQSGOUW","download_json":"https://pith.science/pith/TPND2KRXWWBMNZWLSA4BQSGOUW.json","view_paper":"https://pith.science/paper/TPND2KRX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.09345&json=true","fetch_graph":"https://pith.science/api/pith-number/TPND2KRXWWBMNZWLSA4BQSGOUW/graph.json","fetch_events":"https://pith.science/api/pith-number/TPND2KRXWWBMNZWLSA4BQSGOUW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TPND2KRXWWBMNZWLSA4BQSGOUW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TPND2KRXWWBMNZWLSA4BQSGOUW/action/storage_attestation","attest_author":"https://pith.science/pith/TPND2KRXWWBMNZWLSA4BQSGOUW/action/author_attestation","sign_citation":"https://pith.science/pith/TPND2KRXWWBMNZWLSA4BQSGOUW/action/citation_signature","submit_replication":"https://pith.science/pith/TPND2KRXWWBMNZWLSA4BQSGOUW/action/replication_record"}},"created_at":"2026-07-05T10:48:35.885250+00:00","updated_at":"2026-07-05T10:48:35.885250+00:00"}