{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:T7YOGEZXRPWIL4TCE4DANUROOL","short_pith_number":"pith:T7YOGEZX","schema_version":"1.0","canonical_sha256":"9ff0e313378bec85f262270606d22e72ff353106f9d85c2dcaebd8b33781ab06","source":{"kind":"arxiv","id":"2505.18610","version":1},"attestation_state":"computed","paper":{"title":"PM-KVQ: Progressive Mixed-precision KV Cache Quantization for Long-CoT LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Feng Zhou, Guohao Dai, Huazhong Yang, Jiayi Yang, Shengen Yan, Shiyao Li, Tengxuan Liu, Tianchen Zhao, Xiaohui Song, Yu Wang","submitted_at":"2025-05-24T09:18:11Z","abstract_excerpt":"Recently, significant progress has been made in developing reasoning-capable Large Language Models (LLMs) through long Chain-of-Thought (CoT) techniques. However, this long-CoT reasoning process imposes substantial memory overhead due to the large Key-Value (KV) Cache memory overhead. Post-training KV Cache quantization has emerged as a promising compression technique and has been extensively studied in short-context scenarios. However, directly applying existing methods to long-CoT LLMs causes significant performance degradation due to the following two reasons: (1) Large cumulative error: Ex"},"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":"2505.18610","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-24T09:18:11Z","cross_cats_sorted":[],"title_canon_sha256":"f8179206e1e283a0be5ceb93242cf1a050b910165ea205635cf96e0f27329156","abstract_canon_sha256":"64e49101cf89a07a3ae31a26b9071e87def338ac68ec796f90e18911cb989be8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:09:16.389286Z","signature_b64":"EGTC7KKF5Q9h64ujgzJB6R2v1SJYTwCCPMIVF5Qfryf4Uu7kXjLsDFqKo6sYUGUZSgwGzp9CpO4ANi3AnGJABA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9ff0e313378bec85f262270606d22e72ff353106f9d85c2dcaebd8b33781ab06","last_reissued_at":"2026-07-05T11:09:16.388839Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:09:16.388839Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PM-KVQ: Progressive Mixed-precision KV Cache Quantization for Long-CoT LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Feng Zhou, Guohao Dai, Huazhong Yang, Jiayi Yang, Shengen Yan, Shiyao Li, Tengxuan Liu, Tianchen Zhao, Xiaohui Song, Yu Wang","submitted_at":"2025-05-24T09:18:11Z","abstract_excerpt":"Recently, significant progress has been made in developing reasoning-capable Large Language Models (LLMs) through long Chain-of-Thought (CoT) techniques. However, this long-CoT reasoning process imposes substantial memory overhead due to the large Key-Value (KV) Cache memory overhead. Post-training KV Cache quantization has emerged as a promising compression technique and has been extensively studied in short-context scenarios. However, directly applying existing methods to long-CoT LLMs causes significant performance degradation due to the following two reasons: (1) Large cumulative error: Ex"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.18610","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/2505.18610/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":"2505.18610","created_at":"2026-07-05T11:09:16.388895+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.18610v1","created_at":"2026-07-05T11:09:16.388895+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.18610","created_at":"2026-07-05T11:09:16.388895+00:00"},{"alias_kind":"pith_short_12","alias_value":"T7YOGEZXRPWI","created_at":"2026-07-05T11:09:16.388895+00:00"},{"alias_kind":"pith_short_16","alias_value":"T7YOGEZXRPWIL4TC","created_at":"2026-07-05T11:09:16.388895+00:00"},{"alias_kind":"pith_short_8","alias_value":"T7YOGEZX","created_at":"2026-07-05T11:09:16.388895+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":5,"sample":[{"citing_arxiv_id":"2605.06675","citing_title":"RateQuant: Optimal Mixed-Precision KV Cache Quantization via Rate-Distortion Theory","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2606.24033","citing_title":"RoPE-Aware Bit Allocation for KV-Cache Quantization","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2605.17757","citing_title":"OSCAR: Offline Spectral Covariance-Aware Rotation for 2-bit KV Cache Quantization","ref_index":47,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13734","citing_title":"KVServe: Service-Aware KV Cache Compression for Communication-Efficient Disaggregated LLM Serving","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06675","citing_title":"RateQuant: Optimal Mixed-Precision KV Cache Quantization via Rate-Distortion Theory","ref_index":3,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T7YOGEZXRPWIL4TCE4DANUROOL","json":"https://pith.science/pith/T7YOGEZXRPWIL4TCE4DANUROOL.json","graph_json":"https://pith.science/api/pith-number/T7YOGEZXRPWIL4TCE4DANUROOL/graph.json","events_json":"https://pith.science/api/pith-number/T7YOGEZXRPWIL4TCE4DANUROOL/events.json","paper":"https://pith.science/paper/T7YOGEZX"},"agent_actions":{"view_html":"https://pith.science/pith/T7YOGEZXRPWIL4TCE4DANUROOL","download_json":"https://pith.science/pith/T7YOGEZXRPWIL4TCE4DANUROOL.json","view_paper":"https://pith.science/paper/T7YOGEZX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.18610&json=true","fetch_graph":"https://pith.science/api/pith-number/T7YOGEZXRPWIL4TCE4DANUROOL/graph.json","fetch_events":"https://pith.science/api/pith-number/T7YOGEZXRPWIL4TCE4DANUROOL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T7YOGEZXRPWIL4TCE4DANUROOL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T7YOGEZXRPWIL4TCE4DANUROOL/action/storage_attestation","attest_author":"https://pith.science/pith/T7YOGEZXRPWIL4TCE4DANUROOL/action/author_attestation","sign_citation":"https://pith.science/pith/T7YOGEZXRPWIL4TCE4DANUROOL/action/citation_signature","submit_replication":"https://pith.science/pith/T7YOGEZXRPWIL4TCE4DANUROOL/action/replication_record"}},"created_at":"2026-07-05T11:09:16.388895+00:00","updated_at":"2026-07-05T11:09:16.388895+00:00"}