{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:DKY2UYYPFDTU3NBPWKIIRI7FWK","short_pith_number":"pith:DKY2UYYP","schema_version":"1.0","canonical_sha256":"1ab1aa630f28e74db42fb29088a3e5b2be7dcc53b59ae0b4b6fac4f6510c5bd4","source":{"kind":"arxiv","id":"2502.00527","version":2},"attestation_state":"computed","paper":{"title":"PolarQuant: Leveraging Polar Transformation for Efficient Key Cache Quantization and Decoding Acceleration","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Ang Lv, Guojun Yin, Rui Yan, Songhao Wu, Wei Lin, Xiao Feng, Xun Zhang, Yufei Zhang","submitted_at":"2025-02-01T18:59:03Z","abstract_excerpt":"The KV cache in large language models is a dominant factor in memory usage, limiting their broader applicability. Quantizing the cache to lower bit widths is an effective way to reduce computational costs; however, previous methods struggle with quantizing key vectors due to outliers, resulting in excessive overhead. We propose a novel quantization approach called PolarQuant, which efficiently addresses the outlier challenge. We observe that outliers typically appear in only one of two dimensions, which are rotated together by a specific angle when rotary position embeddings are applied. When "},"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":"2502.00527","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-02-01T18:59:03Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"ade1361c59b6b270ae98b608cb0538ab71945e04955a973f4c14af9360e8b625","abstract_canon_sha256":"e3fba346867a0317a9046dea20d9a95b74571237a13d1761fd5d23841ebc2567"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-08T01:03:43.446498Z","signature_b64":"eAUo4dDcyIWw0U7CxvBIp2Y92JBHMvQNUTs4AB3uUoUDvzTmICamWThJoiiNxcUG8GeNVWPzsZxvR9ad+1sHDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1ab1aa630f28e74db42fb29088a3e5b2be7dcc53b59ae0b4b6fac4f6510c5bd4","last_reissued_at":"2026-06-08T01:03:43.445393Z","signature_status":"signed_v1","first_computed_at":"2026-06-08T01:03:43.445393Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PolarQuant: Leveraging Polar Transformation for Efficient Key Cache Quantization and Decoding Acceleration","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Ang Lv, Guojun Yin, Rui Yan, Songhao Wu, Wei Lin, Xiao Feng, Xun Zhang, Yufei Zhang","submitted_at":"2025-02-01T18:59:03Z","abstract_excerpt":"The KV cache in large language models is a dominant factor in memory usage, limiting their broader applicability. Quantizing the cache to lower bit widths is an effective way to reduce computational costs; however, previous methods struggle with quantizing key vectors due to outliers, resulting in excessive overhead. We propose a novel quantization approach called PolarQuant, which efficiently addresses the outlier challenge. We observe that outliers typically appear in only one of two dimensions, which are rotated together by a specific angle when rotary position embeddings are applied. When "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.00527","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/2502.00527/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":"2502.00527","created_at":"2026-06-08T01:03:43.445545+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.00527v2","created_at":"2026-06-08T01:03:43.445545+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.00527","created_at":"2026-06-08T01:03:43.445545+00:00"},{"alias_kind":"pith_short_12","alias_value":"DKY2UYYPFDTU","created_at":"2026-06-08T01:03:43.445545+00:00"},{"alias_kind":"pith_short_16","alias_value":"DKY2UYYPFDTU3NBP","created_at":"2026-06-08T01:03:43.445545+00:00"},{"alias_kind":"pith_short_8","alias_value":"DKY2UYYP","created_at":"2026-06-08T01:03:43.445545+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.21226","citing_title":"OCTOPUS: Optimized KV Cache for Transformers via Octahedral Parametrization Under optimal Squared error quantization","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18856","citing_title":"SPHERICAL KV: Angle-Domain Attention and Rate-Distortion Retention for Efficient Long-Context Inference","ref_index":24,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DKY2UYYPFDTU3NBPWKIIRI7FWK","json":"https://pith.science/pith/DKY2UYYPFDTU3NBPWKIIRI7FWK.json","graph_json":"https://pith.science/api/pith-number/DKY2UYYPFDTU3NBPWKIIRI7FWK/graph.json","events_json":"https://pith.science/api/pith-number/DKY2UYYPFDTU3NBPWKIIRI7FWK/events.json","paper":"https://pith.science/paper/DKY2UYYP"},"agent_actions":{"view_html":"https://pith.science/pith/DKY2UYYPFDTU3NBPWKIIRI7FWK","download_json":"https://pith.science/pith/DKY2UYYPFDTU3NBPWKIIRI7FWK.json","view_paper":"https://pith.science/paper/DKY2UYYP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.00527&json=true","fetch_graph":"https://pith.science/api/pith-number/DKY2UYYPFDTU3NBPWKIIRI7FWK/graph.json","fetch_events":"https://pith.science/api/pith-number/DKY2UYYPFDTU3NBPWKIIRI7FWK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DKY2UYYPFDTU3NBPWKIIRI7FWK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DKY2UYYPFDTU3NBPWKIIRI7FWK/action/storage_attestation","attest_author":"https://pith.science/pith/DKY2UYYPFDTU3NBPWKIIRI7FWK/action/author_attestation","sign_citation":"https://pith.science/pith/DKY2UYYPFDTU3NBPWKIIRI7FWK/action/citation_signature","submit_replication":"https://pith.science/pith/DKY2UYYPFDTU3NBPWKIIRI7FWK/action/replication_record"}},"created_at":"2026-06-08T01:03:43.445545+00:00","updated_at":"2026-06-08T01:03:43.445545+00:00"}