{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:HKNIHO4QFLZG2VZKX2YRGXOCFR","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":"83b6017948d1ddfa7fe6d32c85acbd709353e81c36033790db0a1379dd017175","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-03-24T07:58:42Z","title_canon_sha256":"0237fd64785f5c9872c9b505fb8aae054d36de633f147c5171a02e95736711a1"},"schema_version":"1.0","source":{"id":"2603.22910","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2603.22910","created_at":"2026-05-18T03:09:22Z"},{"alias_kind":"arxiv_version","alias_value":"2603.22910v2","created_at":"2026-05-18T03:09:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.22910","created_at":"2026-05-18T03:09:22Z"},{"alias_kind":"pith_short_12","alias_value":"HKNIHO4QFLZG","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"HKNIHO4QFLZG2VZK","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"HKNIHO4Q","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:62c3cc1468d5a325f9bdd3a805e158c35db4e5dcba09b1eb9e252b6b7cccfa5a","target":"graph","created_at":"2026-05-18T03:09:22Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"EchoKV consistently outperforms existing methods across multiple compression ratios and backbone models while preserving the throughput of full-cache inference in short-context scenarios."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That intrinsic inter-layer and intra-layer similarities among attention heads are sufficiently stable and informative for a lightweight network to accurately reconstruct the discarded KV components without introducing errors that degrade downstream performance."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"EchoKV compresses LLM KV caches by reconstructing missing components from partial data via inter- and intra-layer attention similarities, outperforming prior methods on LongBench and RULER while supporting on-demand full-cache inference."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"EchoKV compresses the KV cache by reconstructing discarded components from retained ones using attention head similarities."}],"snapshot_sha256":"6fda05209f4ea21a0fa3febe88df0a5409a9e2c91b2bdf0973c4c6b145296c55"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"55a2138617dd87589c4dacb9ca17f9f7c2ba662dcb063e4dc87ac0a0f97a0c17"},"paper":{"abstract_excerpt":"The increasing memory demand of the Key-Value (KV) cache poses a significant bottleneck for Large Language Models (LLMs) in long-context applications. Existing low-rank KV compression methods reduce this footprint by modifying model projections, limiting the flexibility to switch back to standard full-cache inference when sufficient memory is available. In this paper, we propose EchoKV, a flexible KV cache compression framework that supports on-demand transitions from full KV caching to compressed caching. Unlike traditional compression-decompression paradigms, EchoKV utilizes a lightweight ne","authors_text":"Qingfu Zhu, Shiyu Ji, Wanxiang Che, Yijun Liu, Yixuan Wang","cross_cats":[],"headline":"EchoKV compresses the KV cache by reconstructing discarded components from retained ones using attention head similarities.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-03-24T07:58:42Z","title":"EchoKV: Efficient KV Cache Compression via Similarity-Based Reconstruction"},"references":{"count":22,"internal_anchors":11,"resolved_work":22,"sample":[{"cited_arxiv_id":"2305.13245","doi":"","is_internal_anchor":true,"ref_index":1,"title":"GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints","work_id":"b73ad5b2-e553-4c71-b0c9-67e67ba7b158","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"xkv: Cross-layer svd for kv-cache compression","work_id":"16dc86cb-7b66-400c-be1f-dd459db6f94e","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Palu: Compressing kv-cache with low-rank projection.arXiv preprint arXiv:2407.21118","work_id":"dc6247ce-a2da-431d-a935-3fb13543cb13","year":null},{"cited_arxiv_id":"2503.09567","doi":"","is_internal_anchor":true,"ref_index":4,"title":"Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models","work_id":"0b361fed-cf2a-4b90-b61a-de88de4b8840","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Homogeneous keys, heterogeneous values: Exploiting local kv cache asymmetry for long-context llms.arXiv preprint arXiv:2506.05410. Tri Dao","work_id":"4b5ebee2-c5a9-4d4e-a150-51b305167091","year":null}],"snapshot_sha256":"037438b6a30a26a8fe206b2768d274d7802774da5a189c9a422e31b3bd371606"},"source":{"id":"2603.22910","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-15T01:05:08.778242Z","id":"09112ba9-99ed-4b8b-a8e0-b869e68825d5","model_set":{"reader":"grok-4.3"},"one_line_summary":"EchoKV compresses LLM KV caches by reconstructing missing components from partial data via inter- and intra-layer attention similarities, outperforming prior methods on LongBench and RULER while supporting on-demand full-cache inference.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"EchoKV compresses the KV cache by reconstructing discarded components from retained ones using attention head similarities.","strongest_claim":"EchoKV consistently outperforms existing methods across multiple compression ratios and backbone models while preserving the throughput of full-cache inference in short-context scenarios.","weakest_assumption":"That intrinsic inter-layer and intra-layer similarities among attention heads are sufficiently stable and informative for a lightweight network to accurately reconstruct the discarded KV components without introducing errors that degrade downstream performance."}},"verdict_id":"09112ba9-99ed-4b8b-a8e0-b869e68825d5"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:b482d57e8ae41eb2ff3f15a95ca80954cd77e63e9bcd508e28940feb34b02d14","target":"record","created_at":"2026-05-18T03:09:22Z","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":"83b6017948d1ddfa7fe6d32c85acbd709353e81c36033790db0a1379dd017175","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-03-24T07:58:42Z","title_canon_sha256":"0237fd64785f5c9872c9b505fb8aae054d36de633f147c5171a02e95736711a1"},"schema_version":"1.0","source":{"id":"2603.22910","kind":"arxiv","version":2}},"canonical_sha256":"3a9a83bb902af26d572abeb1135dc22c45a496fe6d2f1aed316d8d677dd3a4a6","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3a9a83bb902af26d572abeb1135dc22c45a496fe6d2f1aed316d8d677dd3a4a6","first_computed_at":"2026-05-18T03:09:22.581986Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:09:22.581986Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"VCY0D5cElCr4UBL6gtDp8fvwU47kWZc9w318/qI246wjskNSWqE8aOjDLKrmfTDRHxmdhm7W7LIwB32hdKrtCA==","signature_status":"signed_v1","signed_at":"2026-05-18T03:09:22.582638Z","signed_message":"canonical_sha256_bytes"},"source_id":"2603.22910","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b482d57e8ae41eb2ff3f15a95ca80954cd77e63e9bcd508e28940feb34b02d14","sha256:62c3cc1468d5a325f9bdd3a805e158c35db4e5dcba09b1eb9e252b6b7cccfa5a"],"state_sha256":"751baa3a0d5f2c13c5bb8da804a2703af5d7fdf561a8dcf8f817b48de4c05210"}