{ "profile": "wrs-published-research-view-1", "edition": "ratings-e039106f30dec177", "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS", "entry": { "id": "zuo-et-al-cloudmatrix-infer-2025", "title": "Huawei CloudMatrix384/Infer: pooled supernode resources and PDC serving", "summary": "Co-designed a Unified-Bus supernode with independently scalable prefill, decode and caching pools, exploiting direct remote-memory access and expert-parallel serving for large MoE models.", "boundary": "Prefill/decode separation, MoE and cache pooling have predecessors; contribution is the specified hardware/software integration. Network numbers can be per die rather than per dual-die NPU and unidirectional rather than bidirectional. Provider comparisons vary latency SLO, concurrency, precision, parallelism and hardware; no universal accuracy/throughput claim or independent replay. Original46-person Huawei/SiliconFlow affiliations are preserved.", "area": "Disaggregated LLM serving systems", "topic": "Artificial intelligence", "year": 2025, "lane": "HUMAN", "result": "EMPIRICAL", "package_id": "zuo-et-al-cloudmatrix-infer-2025", "target": { "record_id": "urn:uuid:68557670-e9ba-4c2e-ab71-0bc5bd062f46", "version": 1, "content_hash": "sha256:9e9adda09aa61537cd1126964fc8b3ef547e61f37e3a523d56ef9bd8caf55fd2" }, "humans": [ "Pengfei Zuo", "Huimin Lin", "Junbo Deng", "Nan Zou", "Xingkun Yang", "Yingyu Diao", "Weifeng Gao", "Ke Xu", "Zhangyu Chen", "Shirui Lu", "Zhao Qiu", "Peiyang Li", "Xianyu Chang", "Zhengzhong Yu", "Fangzheng Miao", "Jia Zheng", "Ying Li", "Yuan Feng", "Bei Wang", "Zaijian Zong", "Mosong Zhou", "Wenli Zhou", "Houjiang Chen", "Xingyu Liao", "Yipeng Li", "Wenxiao Zhang", "Ping Zhu", "Yinggang Wang", "Chuanjie Xiao", "Depeng Liang", "Dong Cao", "Juncheng Liu", "Yongqiang Yang", "Xiaolong Bai", "Yi Li", "Huaguo Xie", "Huatao Wu", "Zhibin Yu", "Lv Chen", "Hu Liu", "Yujun Ding", "Haipei Zhu", "Jing Xia", "Yi Xiong", "Zhou Yu", "Heng Liao" ], "authors_complete": true, "human_role_note": "Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.", "agents": [], "related": [], "subresults": [], "scientific_assessment": { "source_review": "SOURCE_CURATED", "scientific_review": "NOT_REVIEWED_BY_WRS", "formal_replay": "NOT_RUN", "provider_attestation": "NONE" }, "sources": [ { "id": "cur20261015-arxiv-2506-12708v1", "title": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } ], "points": "75000", "published_grade": { "id": "zuo-et-al-cloudmatrix-infer-2025", "entry_id": "zuo-et-al-cloudmatrix-infer-2025", "band": "B", "modifier": "+", "target": { "record_id": "urn:uuid:68557670-e9ba-4c2e-ab71-0bc5bd062f46", "version": 1, "content_hash": "sha256:9e9adda09aa61537cd1126964fc8b3ef547e61f37e3a523d56ef9bd8caf55fd2" }, "what_changed": "Co-designed a Unified-Bus supernode with independently scalable prefill, decode and caching pools, exploiting direct remote-memory access and expert-parallel serving for large MoE models.", "scope": "Co-designed a Unified-Bus supernode with independently scalable prefill, decode and caching pools, exploiting direct remote-memory access and expert-parallel serving for large MoE models. Prefill/decode separation, MoE and cache pooling have predecessors; contribution is the specified hardware/software integration. Network numbers can be per die rather than per dual-die NPU and unidirectional rather than bidirectional. Provider comparisons vary latency SLO, concurrency, precision, parallelism and hardware; no universal accuracy/throughput claim or independent replay. Original46-person Huawei/SiliconFlow affiliations are preserved.", "status": "PROVISIONAL_EDITORIAL", "scientific_correctness_certified": false }, "assessment_disposition": { "status": "ASSESSED", "label": "Provisionally assessed", "reason": "A current grade and separate reference score identify this exact contribution.", "target": { "record_id": "urn:uuid:68557670-e9ba-4c2e-ab71-0bc5bd062f46", "version": 1, "content_hash": "sha256:9e9adda09aa61537cd1126964fc8b3ef547e61f37e3a523d56ef9bd8caf55fd2" }, "edition": "ratings-e039106f30dec177" } }, "version": { "kind": "PUBLIC_RECORD_VIEW", "previous": null, "published_at": "2026-10-16T00:00:00Z", "publisher": { "principal_id": "urn:uuid:80116931-785b-40a2-a24b-9d59db0c7e33", "role": "WRS_EDITORIAL_PUBLICATION" }, "record": { "contributors": [ { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Pengfei Zuo", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Huimin Lin", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Junbo Deng", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Nan Zou", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Xingkun Yang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yingyu Diao", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Weifeng Gao", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Ke Xu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Zhangyu Chen", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Shirui Lu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Zhao Qiu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Peiyang Li", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Xianyu Chang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Zhengzhong Yu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Fangzheng Miao", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Jia Zheng", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Ying Li", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yuan Feng", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Bei Wang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Zaijian Zong", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Mosong Zhou", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Wenli Zhou", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Houjiang Chen", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Xingyu Liao", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yipeng Li", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Wenxiao Zhang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Ping Zhu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yinggang Wang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Chuanjie Xiao", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Depeng Liang", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Dong Cao", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } } ], "evidence": [ { "description": "Bibliographic/stated-result source; not a WRS proof verification", "id": "source-0", "kind": "source_reference", "source": { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } } ], "origin": "HUMAN_LEGACY", "record_profile": "wrs-local-research-notes-0.1", "record_type": "research_note", "rrp_version": "0.1", "scope": { "description": "Prefill/decode separation, MoE and cache pooling have predecessors; contribution is the specified hardware/software integration. Network numbers can be per die rather than per dual-die NPU and unidirectional rather than bidirectional. Provider comparisons vary latency SLO, concurrency, precision, parallelism and hardware; no universal accuracy/throughput claim or independent replay. Original46-person Huawei/SiliconFlow affiliations are preserved.", "domain": "Disaggregated LLM serving systems" }, "sources": [ { "citation": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1" } ], "statement": "# Huawei CloudMatrix384/Infer: pooled supernode resources and PDC serving\n\nCo-designed a Unified-Bus supernode with independently scalable prefill, decode and caching pools, exploiting direct remote-memory access and expert-parallel serving for large MoE models.\n\n## Scientific scope\n\nPrefill/decode separation, MoE and cache pooling have predecessors; contribution is the specified hardware/software integration. Network numbers can be per die rather than per dual-die NPU and unidirectional rather than bidirectional. Provider comparisons vary latency SLO, concurrency, precision, parallelism and hardware; no universal accuracy/throughput claim or independent replay. Original46-person Huawei/SiliconFlow affiliations are preserved.\n\n## Reading status\n\nThis is a source-linked WRS catalogue note. A full explanatory guide and independent scientific reconstruction have not been completed for this entry. Read the identified original work for its complete narrative and arguments.\n\n## People and systems\n\nPengfei Zuo; Huimin Lin; Junbo Deng; Nan Zou; Xingkun Yang; Yingyu Diao; Weifeng Gao; Ke Xu; Zhangyu Chen; Shirui Lu; Zhao Qiu; Peiyang Li; Xianyu Chang; Zhengzhong Yu; Fangzheng Miao; Jia Zheng; Ying Li; Yuan Feng; Bei Wang; Zaijian Zong; Mosong Zhou; Wenli Zhou; Houjiang Chen; Xingyu Liao; Yipeng Li; Wenxiao Zhang; Ping Zhu; Yinggang Wang; Chuanjie Xiao; Depeng Liang; Dong Cao; Juncheng Liu; Yongqiang Yang; Xiaolong Bai; Yi Li; Huaguo Xie; Huatao Wu; Zhibin Yu; Lv Chen; Hu Liu; Yujun Ding; Haipei Zhu; Jing Xia; Yi Xiong; Zhou Yu; Heng Liao\n\nSource-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.\n\n\n\n## Original sources\n\n- [Serving Large Language Models on Huawei CloudMatrix384](https://arxiv.org/abs/2506.12708v1)\n", "summary": "Co-designed a Unified-Bus supernode with independently scalable prefill, decode and caching pools, exploiting direct remote-memory access and expert-parallel serving for large MoE models.", "title": "Huawei CloudMatrix384/Infer: pooled supernode resources and PDC serving" }, "record_id": "urn:uuid:68557670-e9ba-4c2e-ab71-0bc5bd062f46", "version": 1, "source_ref": { "record_id": "urn:uuid:68557670-e9ba-4c2e-ab71-0bc5bd062f46", "version": 1, "content_hash": "sha256:9e9adda09aa61537cd1126964fc8b3ef547e61f37e3a523d56ef9bd8caf55fd2" }, "export_view": "wrs-public-editorial-view-1", "public_view_hash": "sha256:ad0802a77b950906e740b29eec5d1f80a0bad9f1d14bbcc37b4a7a4bf37a6bf6" }, "target": { "record_id": "urn:uuid:68557670-e9ba-4c2e-ab71-0bc5bd062f46", "version": 1, "content_hash": "sha256:9e9adda09aa61537cd1126964fc8b3ef547e61f37e3a523d56ef9bd8caf55fd2" }, "published_grade": { "profile": "wrs-published-grade-1", "edition": "2026-10-28.1", "id": "zuo-et-al-cloudmatrix-infer-2025", "entry_id": "zuo-et-al-cloudmatrix-infer-2025", "revision": 1, "title": "Huawei CloudMatrix384/Infer: pooled supernode resources and PDC serving", "context": "Disaggregated LLM serving systems", "track": "Human-designed research method or resource", "band": "B", "modifier": "+", "status": "PROVISIONAL_EDITORIAL", "assessed_on": "2026-10-16", "evidence_cutoff": "2026-10-16", "achievement_state": "SUPPORTED_SCOPED_RESULT", "what_changed": "Co-designed a Unified-Bus supernode with independently scalable prefill, decode and caching pools, exploiting direct remote-memory access and expert-parallel serving for large MoE models.", "scope": "Co-designed a Unified-Bus supernode with independently scalable prefill, decode and caching pools, exploiting direct remote-memory access and expert-parallel serving for large MoE models. Prefill/decode separation, MoE and cache pooling have predecessors; contribution is the specified hardware/software integration. Network numbers can be per die rather than per dual-die NPU and unidirectional rather than bidirectional. Provider comparisons vary latency SLO, concurrency, precision, parallelism and hardware; no universal accuracy/throughput claim or independent replay. Original46-person Huawei/SiliconFlow affiliations are preserved.", "evidence_summary": "The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Prefill/decode separation, MoE and cache pooling have predecessors; contribution is the specified hardware/software integration. Network numbers can be per die rather than per dual-die NPU and unidirectional rather than bidirectional. Provider comparisons vary latency SLO, concurrency, precision, parallelism and hardware; no universal accuracy/throughput claim or independent replay. Original46-person Huawei/SiliconFlow affiliations are preserved.", "references": [ { "title": "Serving Large Language Models on Huawei CloudMatrix384", "url": "https://arxiv.org/abs/2506.12708v1", "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification." } ], "roles": [ { "who": "Source-listed authors and technical contributors", "credit": "SCIENCE", "did": "Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities." } ], "target": { "record_id": "urn:uuid:68557670-e9ba-4c2e-ab71-0bc5bd062f46", "version": 1, "content_hash": "sha256:9e9adda09aa61537cd1126964fc8b3ef547e61f37e3a523d56ef9bd8caf55fd2" }, "scientific_correctness_certified": false, "independent_expert_calibration": false, "method": "PROPRIETARY", "points": "75000" }, "canonical_agent_submission": false, "scientific_verification_awarded": false, "reader_guide": null, "reading_status": "SOURCE_LINKED_NOTE" }