{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:SLKSCT6XWDRZFMSCLQU6MKDNRB","short_pith_number":"pith:SLKSCT6X","schema_version":"1.0","canonical_sha256":"92d5214fd7b0e392b2425c29e6286d884b1fda05d7aa19d0a85c317434b27671","source":{"kind":"arxiv","id":"2312.12456","version":2},"attestation_state":"computed","paper":{"title":"PowerInfer: Fast Large Language Model Serving with a Consumer-grade GPU","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.OS"],"primary_cat":"cs.LG","authors_text":"Haibo Chen, Haotong Xie, Yixin Song, Zeyu Mi","submitted_at":"2023-12-16T02:27:00Z","abstract_excerpt":"This paper introduces PowerInfer, a high-speed Large Language Model (LLM) inference engine on a personal computer (PC) equipped with a single consumer-grade GPU. The key principle underlying the design of PowerInfer is exploiting the high locality inherent in LLM inference, characterized by a power-law distribution in neuron activation. This distribution indicates that a small subset of neurons, termed hot neurons, are consistently activated across inputs, while the majority, cold neurons, vary based on specific inputs. PowerInfer exploits such an insight to design a GPU-CPU hybrid inference e"},"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":"2312.12456","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-12-16T02:27:00Z","cross_cats_sorted":["cs.OS"],"title_canon_sha256":"abbbbdc2a11407a4f8d33f93b816db8cb91ae14ba66be8ef300a08230beca6c6","abstract_canon_sha256":"55a761ff35471727ff3c67504a1c4b19c4d8e8b1d71ae49bfca54564f5d39528"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:48:11.633863Z","signature_b64":"yX8D2h3qwYaIZu9oRDSgvss2Y/FYP0IIS6h7LIOUYz109WspjoYOf1Qg/tKmmQk2putmDU8QdaYM6srmzl0HDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"92d5214fd7b0e392b2425c29e6286d884b1fda05d7aa19d0a85c317434b27671","last_reissued_at":"2026-07-05T09:48:11.633438Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:48:11.633438Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PowerInfer: Fast Large Language Model Serving with a Consumer-grade GPU","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.OS"],"primary_cat":"cs.LG","authors_text":"Haibo Chen, Haotong Xie, Yixin Song, Zeyu Mi","submitted_at":"2023-12-16T02:27:00Z","abstract_excerpt":"This paper introduces PowerInfer, a high-speed Large Language Model (LLM) inference engine on a personal computer (PC) equipped with a single consumer-grade GPU. The key principle underlying the design of PowerInfer is exploiting the high locality inherent in LLM inference, characterized by a power-law distribution in neuron activation. This distribution indicates that a small subset of neurons, termed hot neurons, are consistently activated across inputs, while the majority, cold neurons, vary based on specific inputs. PowerInfer exploits such an insight to design a GPU-CPU hybrid inference e"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2312.12456","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/2312.12456/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":"2312.12456","created_at":"2026-07-05T09:48:11.633497+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.12456v2","created_at":"2026-07-05T09:48:11.633497+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.12456","created_at":"2026-07-05T09:48:11.633497+00:00"},{"alias_kind":"pith_short_12","alias_value":"SLKSCT6XWDRZ","created_at":"2026-07-05T09:48:11.633497+00:00"},{"alias_kind":"pith_short_16","alias_value":"SLKSCT6XWDRZFMSC","created_at":"2026-07-05T09:48:11.633497+00:00"},{"alias_kind":"pith_short_8","alias_value":"SLKSCT6X","created_at":"2026-07-05T09:48:11.633497+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21428","citing_title":"Does Mixture-of-Experts Actually Help Inference on Consumer and Edge Hardware? An Empirical Study","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12556","citing_title":"ITME: Inference Tiered Memory Expansion with Disaggregated CXL-Hybrid Memories","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27476","citing_title":"EdgeFM: Efficient Edge Inference for Vision-Language Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2401.05459","citing_title":"Personal LLM Agents: Insights and Survey about the Capability, Efficiency and Security","ref_index":274,"is_internal_anchor":false},{"citing_arxiv_id":"2404.14294","citing_title":"A Survey on Efficient Inference for Large Language Models","ref_index":251,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27476","citing_title":"EdgeFM: Efficient Edge Inference for Vision-Language Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27115","citing_title":"Exploring the Limits of Pruning: Task-Specific Neurons, Model Collapse, and Recovery in Task-Specific Large Language Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2409.19256","citing_title":"HybridFlow: A Flexible and Efficient RLHF Framework","ref_index":83,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17182","citing_title":"Layer-wise MoE Routing Locality under Shared-Prefix Code Generation: Token-Identity Decomposition and Compile-Equivalent Fork Redundancy","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SLKSCT6XWDRZFMSCLQU6MKDNRB","json":"https://pith.science/pith/SLKSCT6XWDRZFMSCLQU6MKDNRB.json","graph_json":"https://pith.science/api/pith-number/SLKSCT6XWDRZFMSCLQU6MKDNRB/graph.json","events_json":"https://pith.science/api/pith-number/SLKSCT6XWDRZFMSCLQU6MKDNRB/events.json","paper":"https://pith.science/paper/SLKSCT6X"},"agent_actions":{"view_html":"https://pith.science/pith/SLKSCT6XWDRZFMSCLQU6MKDNRB","download_json":"https://pith.science/pith/SLKSCT6XWDRZFMSCLQU6MKDNRB.json","view_paper":"https://pith.science/paper/SLKSCT6X","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.12456&json=true","fetch_graph":"https://pith.science/api/pith-number/SLKSCT6XWDRZFMSCLQU6MKDNRB/graph.json","fetch_events":"https://pith.science/api/pith-number/SLKSCT6XWDRZFMSCLQU6MKDNRB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SLKSCT6XWDRZFMSCLQU6MKDNRB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SLKSCT6XWDRZFMSCLQU6MKDNRB/action/storage_attestation","attest_author":"https://pith.science/pith/SLKSCT6XWDRZFMSCLQU6MKDNRB/action/author_attestation","sign_citation":"https://pith.science/pith/SLKSCT6XWDRZFMSCLQU6MKDNRB/action/citation_signature","submit_replication":"https://pith.science/pith/SLKSCT6XWDRZFMSCLQU6MKDNRB/action/replication_record"}},"created_at":"2026-07-05T09:48:11.633497+00:00","updated_at":"2026-07-05T09:48:11.633497+00:00"}