{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:6XXT4URYLA3LSH6XRRAAGVIYBK","short_pith_number":"pith:6XXT4URY","schema_version":"1.0","canonical_sha256":"f5ef3e52385836b91fd78c400355180a8cce2956996a0cb1918697fd5259c152","source":{"kind":"arxiv","id":"2509.03136","version":1},"attestation_state":"computed","paper":{"title":"Adaptive KV-Cache Compression without Manually Setting Budget","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.DB","authors_text":"Chenxia Tang, Hongli Xu, Jianchun Liu, Liusheng Huang","submitted_at":"2025-09-03T08:38:40Z","abstract_excerpt":"Large language models (LLMs) inference relies heavily on KV-caches to accelerate autoregressive decoding, but the resulting memory footprint grows rapidly with sequence length, posing significant efficiency challenges. Current KV-cache compression methods suffer from a Procrustes' bed problem: they force diverse workloads into fixed compression ratios, leading to suboptimal resource allocation and inference performance. To this end, we present GVote, an adaptive KV-cache compression scheme that eliminates manual budget specification while achieving superior accuracy-efficiency trade-offs. GVot"},"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":"2509.03136","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DB","submitted_at":"2025-09-03T08:38:40Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"5c9d3aaa7e6b25db298987ca91aa551a35e632480bb8d9cc4b790ae9d381a0bf","abstract_canon_sha256":"d245567694e8e5dd8ef80d2419330022e5712b72ee78db74a3941bf60c7bbf93"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:04:04.789258Z","signature_b64":"1LesdvWNux8ogIkvWjCYOpkZ9URPWaee2ZtkAS1DxblJpzgNsWfztUitavBfFckck2blXIoHvSbca8GZTltEDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f5ef3e52385836b91fd78c400355180a8cce2956996a0cb1918697fd5259c152","last_reissued_at":"2026-07-05T12:04:04.788777Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:04:04.788777Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Adaptive KV-Cache Compression without Manually Setting Budget","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.DB","authors_text":"Chenxia Tang, Hongli Xu, Jianchun Liu, Liusheng Huang","submitted_at":"2025-09-03T08:38:40Z","abstract_excerpt":"Large language models (LLMs) inference relies heavily on KV-caches to accelerate autoregressive decoding, but the resulting memory footprint grows rapidly with sequence length, posing significant efficiency challenges. Current KV-cache compression methods suffer from a Procrustes' bed problem: they force diverse workloads into fixed compression ratios, leading to suboptimal resource allocation and inference performance. To this end, we present GVote, an adaptive KV-cache compression scheme that eliminates manual budget specification while achieving superior accuracy-efficiency trade-offs. GVot"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.03136","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/2509.03136/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":"2509.03136","created_at":"2026-07-05T12:04:04.788830+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.03136v1","created_at":"2026-07-05T12:04:04.788830+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.03136","created_at":"2026-07-05T12:04:04.788830+00:00"},{"alias_kind":"pith_short_12","alias_value":"6XXT4URYLA3L","created_at":"2026-07-05T12:04:04.788830+00:00"},{"alias_kind":"pith_short_16","alias_value":"6XXT4URYLA3LSH6X","created_at":"2026-07-05T12:04:04.788830+00:00"},{"alias_kind":"pith_short_8","alias_value":"6XXT4URY","created_at":"2026-07-05T12:04:04.788830+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2602.08686","citing_title":"CompilerKV: Risk-Adaptive KV Compression via Offline Experience Compilation","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6XXT4URYLA3LSH6XRRAAGVIYBK","json":"https://pith.science/pith/6XXT4URYLA3LSH6XRRAAGVIYBK.json","graph_json":"https://pith.science/api/pith-number/6XXT4URYLA3LSH6XRRAAGVIYBK/graph.json","events_json":"https://pith.science/api/pith-number/6XXT4URYLA3LSH6XRRAAGVIYBK/events.json","paper":"https://pith.science/paper/6XXT4URY"},"agent_actions":{"view_html":"https://pith.science/pith/6XXT4URYLA3LSH6XRRAAGVIYBK","download_json":"https://pith.science/pith/6XXT4URYLA3LSH6XRRAAGVIYBK.json","view_paper":"https://pith.science/paper/6XXT4URY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.03136&json=true","fetch_graph":"https://pith.science/api/pith-number/6XXT4URYLA3LSH6XRRAAGVIYBK/graph.json","fetch_events":"https://pith.science/api/pith-number/6XXT4URYLA3LSH6XRRAAGVIYBK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6XXT4URYLA3LSH6XRRAAGVIYBK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6XXT4URYLA3LSH6XRRAAGVIYBK/action/storage_attestation","attest_author":"https://pith.science/pith/6XXT4URYLA3LSH6XRRAAGVIYBK/action/author_attestation","sign_citation":"https://pith.science/pith/6XXT4URYLA3LSH6XRRAAGVIYBK/action/citation_signature","submit_replication":"https://pith.science/pith/6XXT4URYLA3LSH6XRRAAGVIYBK/action/replication_record"}},"created_at":"2026-07-05T12:04:04.788830+00:00","updated_at":"2026-07-05T12:04:04.788830+00:00"}