{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LPV2FZOAL5MQPY2PKNMXTZIH7I","short_pith_number":"pith:LPV2FZOA","schema_version":"1.0","canonical_sha256":"5beba2e5c05f5907e34f535979e507fa20890ee0be7c6426ba4eaae7b7639973","source":{"kind":"arxiv","id":"2502.16886","version":4},"attestation_state":"computed","paper":{"title":"ReFreeKV: Towards Threshold-Free KV Cache Compression","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chenyang Lyu, Fandong Meng, Jie Zhou, Lemao Liu, Liyan Xu, Longyue Wang, Mo Yu, Piji Li, Xuanfan Ni","submitted_at":"2025-02-24T06:33:39Z","abstract_excerpt":"To reduce memory consumption during LLM inference, a handful of methods have been proposed for KV cache pruning. While these techniques can accomplish lossless memory reduction on many datasets, they often hinge on an under-emphasized condition: an input/domain-specific threshold for KV cache budget needs to be pre-determined to achieve the optimal performance. However, such input-sensitive design may be considerably limited in real-world scenarios, as open-domain inputs span diverse domains, lengths and difficulty levels, without clear boundaries for threshold selection. As a result, the depe"},"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":"2502.16886","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-24T06:33:39Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"62e54ee868af9c7e6329aaf8b3c1c0a5f95545160dea7321c8bb3233f35e3980","abstract_canon_sha256":"dbb7472a84dbbb400ba2356332a648b72e9b9b263e147f03073bfff409a82580"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-29T01:14:01.667107Z","signature_b64":"BVQnOf4L3MgpNOtlf+bG58513natdBf0VeiXz94uu4puMa1WOJ2dO3/PBd5AvIRkCeftLaJhSRc7c7yozj6AAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5beba2e5c05f5907e34f535979e507fa20890ee0be7c6426ba4eaae7b7639973","last_reissued_at":"2026-06-29T01:14:01.666641Z","signature_status":"signed_v1","first_computed_at":"2026-06-29T01:14:01.666641Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ReFreeKV: Towards Threshold-Free KV Cache Compression","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chenyang Lyu, Fandong Meng, Jie Zhou, Lemao Liu, Liyan Xu, Longyue Wang, Mo Yu, Piji Li, Xuanfan Ni","submitted_at":"2025-02-24T06:33:39Z","abstract_excerpt":"To reduce memory consumption during LLM inference, a handful of methods have been proposed for KV cache pruning. While these techniques can accomplish lossless memory reduction on many datasets, they often hinge on an under-emphasized condition: an input/domain-specific threshold for KV cache budget needs to be pre-determined to achieve the optimal performance. However, such input-sensitive design may be considerably limited in real-world scenarios, as open-domain inputs span diverse domains, lengths and difficulty levels, without clear boundaries for threshold selection. As a result, the depe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.16886","kind":"arxiv","version":4},"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/2502.16886/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":"2502.16886","created_at":"2026-06-29T01:14:01.666701+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.16886v4","created_at":"2026-06-29T01:14:01.666701+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.16886","created_at":"2026-06-29T01:14:01.666701+00:00"},{"alias_kind":"pith_short_12","alias_value":"LPV2FZOAL5MQ","created_at":"2026-06-29T01:14:01.666701+00:00"},{"alias_kind":"pith_short_16","alias_value":"LPV2FZOAL5MQPY2P","created_at":"2026-06-29T01:14:01.666701+00:00"},{"alias_kind":"pith_short_8","alias_value":"LPV2FZOA","created_at":"2026-06-29T01:14:01.666701+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.28831","citing_title":"HARD-KV: Head-Adaptive Regularization for Decoding-time KV Compression","ref_index":28,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LPV2FZOAL5MQPY2PKNMXTZIH7I","json":"https://pith.science/pith/LPV2FZOAL5MQPY2PKNMXTZIH7I.json","graph_json":"https://pith.science/api/pith-number/LPV2FZOAL5MQPY2PKNMXTZIH7I/graph.json","events_json":"https://pith.science/api/pith-number/LPV2FZOAL5MQPY2PKNMXTZIH7I/events.json","paper":"https://pith.science/paper/LPV2FZOA"},"agent_actions":{"view_html":"https://pith.science/pith/LPV2FZOAL5MQPY2PKNMXTZIH7I","download_json":"https://pith.science/pith/LPV2FZOAL5MQPY2PKNMXTZIH7I.json","view_paper":"https://pith.science/paper/LPV2FZOA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.16886&json=true","fetch_graph":"https://pith.science/api/pith-number/LPV2FZOAL5MQPY2PKNMXTZIH7I/graph.json","fetch_events":"https://pith.science/api/pith-number/LPV2FZOAL5MQPY2PKNMXTZIH7I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LPV2FZOAL5MQPY2PKNMXTZIH7I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LPV2FZOAL5MQPY2PKNMXTZIH7I/action/storage_attestation","attest_author":"https://pith.science/pith/LPV2FZOAL5MQPY2PKNMXTZIH7I/action/author_attestation","sign_citation":"https://pith.science/pith/LPV2FZOAL5MQPY2PKNMXTZIH7I/action/citation_signature","submit_replication":"https://pith.science/pith/LPV2FZOAL5MQPY2PKNMXTZIH7I/action/replication_record"}},"created_at":"2026-06-29T01:14:01.666701+00:00","updated_at":"2026-06-29T01:14:01.666701+00:00"}