{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:L3PFNSLVI3POBIKBF2IV5TQKPI","short_pith_number":"pith:L3PFNSLV","canonical_record":{"source":{"id":"2405.12532","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-21T06:46:37Z","cross_cats_sorted":[],"title_canon_sha256":"45c241aad9978b63a70fb3aef896480e9503ebe14ead062ec4cfa055a8b3c27d","abstract_canon_sha256":"2071930b197b676787b6fe6f1f90960b98069d596e3bd49eb5703f9cdf10b2dd"},"schema_version":"1.0"},"canonical_sha256":"5ede56c97546dee0a1412e915ece0a7a021f26004481fc94523f608559534c22","source":{"kind":"arxiv","id":"2405.12532","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2405.12532","created_at":"2026-07-05T08:27:48Z"},{"alias_kind":"arxiv_version","alias_value":"2405.12532v2","created_at":"2026-07-05T08:27:48Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.12532","created_at":"2026-07-05T08:27:48Z"},{"alias_kind":"pith_short_12","alias_value":"L3PFNSLVI3PO","created_at":"2026-07-05T08:27:48Z"},{"alias_kind":"pith_short_16","alias_value":"L3PFNSLVI3POBIKB","created_at":"2026-07-05T08:27:48Z"},{"alias_kind":"pith_short_8","alias_value":"L3PFNSLV","created_at":"2026-07-05T08:27:48Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:L3PFNSLVI3POBIKBF2IV5TQKPI","target":"record","payload":{"canonical_record":{"source":{"id":"2405.12532","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-21T06:46:37Z","cross_cats_sorted":[],"title_canon_sha256":"45c241aad9978b63a70fb3aef896480e9503ebe14ead062ec4cfa055a8b3c27d","abstract_canon_sha256":"2071930b197b676787b6fe6f1f90960b98069d596e3bd49eb5703f9cdf10b2dd"},"schema_version":"1.0"},"canonical_sha256":"5ede56c97546dee0a1412e915ece0a7a021f26004481fc94523f608559534c22","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:27:48.904963Z","signature_b64":"XOPzMnVeFBtaJszu5ZQpTOIlJ8q9EDJJsS0scG6+qxqnSzKli66DPCQXvKCFJOdE1rAXK/J6v8/QeiP1pHZmDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5ede56c97546dee0a1412e915ece0a7a021f26004481fc94523f608559534c22","last_reissued_at":"2026-07-05T08:27:48.904428Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:27:48.904428Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2405.12532","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:27:48Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LthXTjl7o7V4kbferSl8GZHHmkHYNoizAt/pknsLOJgO8A8VBCGhyEz1whH0Rysjqdg8lMbV1tBhEhJ+51OkAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-14T01:37:56.662520Z"},"content_sha256":"516a6e89c8e823f3fd36e78df18cf6e71b799318580734a55de661ce03bb194b","schema_version":"1.0","event_id":"sha256:516a6e89c8e823f3fd36e78df18cf6e71b799318580734a55de661ce03bb194b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:L3PFNSLVI3POBIKBF2IV5TQKPI","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"PyramidInfer: Pyramid KV Cache Compression for High-throughput LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dongjie Yang, Hai Zhao, Shilin Zhang, Xiaodong Han, Yan Gao, Yao Hu","submitted_at":"2024-05-21T06:46:37Z","abstract_excerpt":"Large Language Models (LLMs) have shown remarkable comprehension abilities but face challenges in GPU memory usage during inference, hindering their scalability for real-time applications like chatbots. To accelerate inference, we store computed keys and values (KV cache) in the GPU memory. Existing methods study the KV cache compression to reduce memory by pruning the pre-computed KV cache. However, they neglect the inter-layer dependency between layers and huge memory consumption in pre-computation. To explore these deficiencies, we find that the number of crucial keys and values that influe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.12532","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/2405.12532/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":"2026-07-05T08:27:48Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"BSmzbpV+N7KasFgZakkMcb0m7X0veOKR/F407tErjyfIb+o2b0Xm3njsQ4He7EgpNgp6jIVrUnjR4VtEAaA+Cg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-14T01:37:56.662894Z"},"content_sha256":"ff37f080c955012e898d13485c2246f37210e392e211c06e747e2aa6a2bd15b9","schema_version":"1.0","event_id":"sha256:ff37f080c955012e898d13485c2246f37210e392e211c06e747e2aa6a2bd15b9"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/L3PFNSLVI3POBIKBF2IV5TQKPI/bundle.json","state_url":"https://pith.science/pith/L3PFNSLVI3POBIKBF2IV5TQKPI/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/L3PFNSLVI3POBIKBF2IV5TQKPI/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":"2026-07-14T01:37:56Z","links":{"resolver":"https://pith.science/pith/L3PFNSLVI3POBIKBF2IV5TQKPI","bundle":"https://pith.science/pith/L3PFNSLVI3POBIKBF2IV5TQKPI/bundle.json","state":"https://pith.science/pith/L3PFNSLVI3POBIKBF2IV5TQKPI/state.json","well_known_bundle":"https://pith.science/.well-known/pith/L3PFNSLVI3POBIKBF2IV5TQKPI/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:L3PFNSLVI3POBIKBF2IV5TQKPI","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":"2071930b197b676787b6fe6f1f90960b98069d596e3bd49eb5703f9cdf10b2dd","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-21T06:46:37Z","title_canon_sha256":"45c241aad9978b63a70fb3aef896480e9503ebe14ead062ec4cfa055a8b3c27d"},"schema_version":"1.0","source":{"id":"2405.12532","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2405.12532","created_at":"2026-07-05T08:27:48Z"},{"alias_kind":"arxiv_version","alias_value":"2405.12532v2","created_at":"2026-07-05T08:27:48Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.12532","created_at":"2026-07-05T08:27:48Z"},{"alias_kind":"pith_short_12","alias_value":"L3PFNSLVI3PO","created_at":"2026-07-05T08:27:48Z"},{"alias_kind":"pith_short_16","alias_value":"L3PFNSLVI3POBIKB","created_at":"2026-07-05T08:27:48Z"},{"alias_kind":"pith_short_8","alias_value":"L3PFNSLV","created_at":"2026-07-05T08:27:48Z"}],"graph_snapshots":[{"event_id":"sha256:ff37f080c955012e898d13485c2246f37210e392e211c06e747e2aa6a2bd15b9","target":"graph","created_at":"2026-07-05T08:27:48Z","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/2405.12532/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large Language Models (LLMs) have shown remarkable comprehension abilities but face challenges in GPU memory usage during inference, hindering their scalability for real-time applications like chatbots. To accelerate inference, we store computed keys and values (KV cache) in the GPU memory. Existing methods study the KV cache compression to reduce memory by pruning the pre-computed KV cache. However, they neglect the inter-layer dependency between layers and huge memory consumption in pre-computation. To explore these deficiencies, we find that the number of crucial keys and values that influe","authors_text":"Dongjie Yang, Hai Zhao, Shilin Zhang, Xiaodong Han, Yan Gao, Yao Hu","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-21T06:46:37Z","title":"PyramidInfer: Pyramid KV Cache Compression for High-throughput LLM Inference"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.12532","kind":"arxiv","version":2},"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":"sha256:516a6e89c8e823f3fd36e78df18cf6e71b799318580734a55de661ce03bb194b","target":"record","created_at":"2026-07-05T08:27:48Z","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":"2071930b197b676787b6fe6f1f90960b98069d596e3bd49eb5703f9cdf10b2dd","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-21T06:46:37Z","title_canon_sha256":"45c241aad9978b63a70fb3aef896480e9503ebe14ead062ec4cfa055a8b3c27d"},"schema_version":"1.0","source":{"id":"2405.12532","kind":"arxiv","version":2}},"canonical_sha256":"5ede56c97546dee0a1412e915ece0a7a021f26004481fc94523f608559534c22","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"5ede56c97546dee0a1412e915ece0a7a021f26004481fc94523f608559534c22","first_computed_at":"2026-07-05T08:27:48.904428Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:27:48.904428Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"XOPzMnVeFBtaJszu5ZQpTOIlJ8q9EDJJsS0scG6+qxqnSzKli66DPCQXvKCFJOdE1rAXK/J6v8/QeiP1pHZmDg==","signature_status":"signed_v1","signed_at":"2026-07-05T08:27:48.904963Z","signed_message":"canonical_sha256_bytes"},"source_id":"2405.12532","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:516a6e89c8e823f3fd36e78df18cf6e71b799318580734a55de661ce03bb194b","sha256:ff37f080c955012e898d13485c2246f37210e392e211c06e747e2aa6a2bd15b9"],"state_sha256":"2ecc00a81fde1c52a3bb348f342fab71ba02b6b9bd04cf4f60d1f086ede60a97"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"K8wqUzmgVkw3T0gTplVemzc+6QO4FwrDUtFF+HP7t28N2f5PLSEmcRC2mvzPhemUOrLy4nKDIKIxnaHlmm8mDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-14T01:37:56.665406Z","bundle_sha256":"62750a8f504076976f2fbe2be082f09a6b189f1c9d6c25d1b4c81302f9c2a317"}}