{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:QGJWGCPJ6KZ65FZXABF53H3OQF","short_pith_number":"pith:QGJWGCPJ","canonical_record":{"source":{"id":"2502.12574","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-18T06:26:05Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"9b237a06dae8c3edea9e885284b9d59a9da8db1cfc8499151362b9f8cac4dfdc","abstract_canon_sha256":"7f3a323aa28c15862d85d723d2dfc4b04b8fbb2cbbbcfd4f62c8310b0e14aa01"},"schema_version":"1.0"},"canonical_sha256":"81936309e9f2b3ee9737004bdd9f6e81404fa629439f64eb8e3a62b696652b48","source":{"kind":"arxiv","id":"2502.12574","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.12574","created_at":"2026-07-05T10:15:57Z"},{"alias_kind":"arxiv_version","alias_value":"2502.12574v1","created_at":"2026-07-05T10:15:57Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.12574","created_at":"2026-07-05T10:15:57Z"},{"alias_kind":"pith_short_12","alias_value":"QGJWGCPJ6KZ6","created_at":"2026-07-05T10:15:57Z"},{"alias_kind":"pith_short_16","alias_value":"QGJWGCPJ6KZ65FZX","created_at":"2026-07-05T10:15:57Z"},{"alias_kind":"pith_short_8","alias_value":"QGJWGCPJ","created_at":"2026-07-05T10:15:57Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:QGJWGCPJ6KZ65FZXABF53H3OQF","target":"record","payload":{"canonical_record":{"source":{"id":"2502.12574","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-18T06:26:05Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"9b237a06dae8c3edea9e885284b9d59a9da8db1cfc8499151362b9f8cac4dfdc","abstract_canon_sha256":"7f3a323aa28c15862d85d723d2dfc4b04b8fbb2cbbbcfd4f62c8310b0e14aa01"},"schema_version":"1.0"},"canonical_sha256":"81936309e9f2b3ee9737004bdd9f6e81404fa629439f64eb8e3a62b696652b48","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:15:57.709039Z","signature_b64":"2CTNz50+LC9J6PJ9gXn/JD6OoZaq9j32wz8wttD3t4mTkuHzFRSudhRTnttdD+MqPkEYFHR3fQvGZhePXupnAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"81936309e9f2b3ee9737004bdd9f6e81404fa629439f64eb8e3a62b696652b48","last_reissued_at":"2026-07-05T10:15:57.708548Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:15:57.708548Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2502.12574","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:15:57Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"TXznosrh+s6uBr4rm9g+1yPVI2ePDTQI9Gau5nVZI6LB2rGbilat6IFcIjQDiGh+sXUK6ryxPwrL2QmuQwaRBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T23:46:20.446733Z"},"content_sha256":"7393497f73be8a78f548e2ba75119db781b31b89b924045679c0cf6acb3c7a68","schema_version":"1.0","event_id":"sha256:7393497f73be8a78f548e2ba75119db781b31b89b924045679c0cf6acb3c7a68"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:QGJWGCPJ6KZ65FZXABF53H3OQF","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"HeadInfer: Memory-Efficient LLM Inference by Head-wise Offloading","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Anima Anandkumar, Beidi Chen, Bo Yuan, Cheng Luo, Hanshi Sun, Jiawei Zhao, Jinqi Xiao, Junjie Hu, Wen Xiao, Zefan Cai","submitted_at":"2025-02-18T06:26:05Z","abstract_excerpt":"Transformer-based large language models (LLMs) demonstrate impressive performance in long context generation. Extending the context length has disproportionately shifted the memory footprint of LLMs during inference to the key-value cache (KV cache). In this paper, we propose HEADINFER, which offloads the KV cache to CPU RAM while avoiding the need to fully store the KV cache for any transformer layer on the GPU. HEADINFER employs a fine-grained, head-wise offloading strategy, maintaining only selective attention heads KV cache on the GPU while computing attention output dynamically. Through r"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.12574","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/2502.12574/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:15:57Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"tQQMJMMij/KP1rwAH/rOapoVfQRNnUybcBBy/uAi8/DxGotDMOlSHSXH8KVFFn0KDnH0vkVCUD4SyxixnOSECw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T23:46:20.447351Z"},"content_sha256":"b7e98a164b1861174b5e798e118b81dff3f6cdbba60816c6c2683d2386939c27","schema_version":"1.0","event_id":"sha256:b7e98a164b1861174b5e798e118b81dff3f6cdbba60816c6c2683d2386939c27"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/QGJWGCPJ6KZ65FZXABF53H3OQF/bundle.json","state_url":"https://pith.science/pith/QGJWGCPJ6KZ65FZXABF53H3OQF/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/QGJWGCPJ6KZ65FZXABF53H3OQF/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-09T23:46:20Z","links":{"resolver":"https://pith.science/pith/QGJWGCPJ6KZ65FZXABF53H3OQF","bundle":"https://pith.science/pith/QGJWGCPJ6KZ65FZXABF53H3OQF/bundle.json","state":"https://pith.science/pith/QGJWGCPJ6KZ65FZXABF53H3OQF/state.json","well_known_bundle":"https://pith.science/.well-known/pith/QGJWGCPJ6KZ65FZXABF53H3OQF/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:QGJWGCPJ6KZ65FZXABF53H3OQF","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"7f3a323aa28c15862d85d723d2dfc4b04b8fbb2cbbbcfd4f62c8310b0e14aa01","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-18T06:26:05Z","title_canon_sha256":"9b237a06dae8c3edea9e885284b9d59a9da8db1cfc8499151362b9f8cac4dfdc"},"schema_version":"1.0","source":{"id":"2502.12574","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.12574","created_at":"2026-07-05T10:15:57Z"},{"alias_kind":"arxiv_version","alias_value":"2502.12574v1","created_at":"2026-07-05T10:15:57Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.12574","created_at":"2026-07-05T10:15:57Z"},{"alias_kind":"pith_short_12","alias_value":"QGJWGCPJ6KZ6","created_at":"2026-07-05T10:15:57Z"},{"alias_kind":"pith_short_16","alias_value":"QGJWGCPJ6KZ65FZX","created_at":"2026-07-05T10:15:57Z"},{"alias_kind":"pith_short_8","alias_value":"QGJWGCPJ","created_at":"2026-07-05T10:15:57Z"}],"graph_snapshots":[{"event_id":"sha256:b7e98a164b1861174b5e798e118b81dff3f6cdbba60816c6c2683d2386939c27","target":"graph","created_at":"2026-07-05T10:15:57Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2502.12574/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Transformer-based large language models (LLMs) demonstrate impressive performance in long context generation. Extending the context length has disproportionately shifted the memory footprint of LLMs during inference to the key-value cache (KV cache). In this paper, we propose HEADINFER, which offloads the KV cache to CPU RAM while avoiding the need to fully store the KV cache for any transformer layer on the GPU. HEADINFER employs a fine-grained, head-wise offloading strategy, maintaining only selective attention heads KV cache on the GPU while computing attention output dynamically. Through r","authors_text":"Anima Anandkumar, Beidi Chen, Bo Yuan, Cheng Luo, Hanshi Sun, Jiawei Zhao, Jinqi Xiao, Junjie Hu, Wen Xiao, Zefan Cai","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-18T06:26:05Z","title":"HeadInfer: Memory-Efficient LLM Inference by Head-wise Offloading"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.12574","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:7393497f73be8a78f548e2ba75119db781b31b89b924045679c0cf6acb3c7a68","target":"record","created_at":"2026-07-05T10:15:57Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"7f3a323aa28c15862d85d723d2dfc4b04b8fbb2cbbbcfd4f62c8310b0e14aa01","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-18T06:26:05Z","title_canon_sha256":"9b237a06dae8c3edea9e885284b9d59a9da8db1cfc8499151362b9f8cac4dfdc"},"schema_version":"1.0","source":{"id":"2502.12574","kind":"arxiv","version":1}},"canonical_sha256":"81936309e9f2b3ee9737004bdd9f6e81404fa629439f64eb8e3a62b696652b48","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"81936309e9f2b3ee9737004bdd9f6e81404fa629439f64eb8e3a62b696652b48","first_computed_at":"2026-07-05T10:15:57.708548Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:15:57.708548Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"2CTNz50+LC9J6PJ9gXn/JD6OoZaq9j32wz8wttD3t4mTkuHzFRSudhRTnttdD+MqPkEYFHR3fQvGZhePXupnAg==","signature_status":"signed_v1","signed_at":"2026-07-05T10:15:57.709039Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.12574","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:7393497f73be8a78f548e2ba75119db781b31b89b924045679c0cf6acb3c7a68","sha256:b7e98a164b1861174b5e798e118b81dff3f6cdbba60816c6c2683d2386939c27"],"state_sha256":"09a5c39e1f8a387515ae3940ef4209a45490c51b0536b441d32b56da3b933912"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Q4VycCEvnJBZh6sdSNoUwfEP9BBfCevMwF9Y+1zCuGUPx0lGW54fNfq2+ROOXJyf2bRLGc6gQeWNlRWOcxaqAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-09T23:46:20.452276Z","bundle_sha256":"a16cd7552d131773e6c447b8d12205b89e31e6605d9208c03adfaeefcdfeb580"}}