{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:M2GCKZZUNZSWZG7ZLUCZOGHE6I","short_pith_number":"pith:M2GCKZZU","schema_version":"1.0","canonical_sha256":"668c2567346e656c9bf95d059718e4f22647b0e783a9fd2b294e477097b76b87","source":{"kind":"arxiv","id":"2507.11273","version":1},"attestation_state":"computed","paper":{"title":"KV-Latent: Dimensional-level KV Cache Reduction with Frequency-aware Rotary Positional Embedding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Baoyuan Qi, Guoming Liu, Hai Zhao, Lefei Zhang, Luohe Shi, Zuchao Li","submitted_at":"2025-07-15T12:52:12Z","abstract_excerpt":"Large language models (LLMs) based on Transformer Decoders have become the preferred choice for conversational generative AI. Despite the overall superiority of the Decoder architecture, the gradually increasing Key-Value (KV) cache during inference has emerged as a primary efficiency bottleneck, both in aspects of memory consumption and data transfer bandwidth limitations. To address these challenges, we propose a paradigm called KV-Latent. By down-sampling the Key-Value vector dimensions into a latent space, we can significantly reduce the KV Cache footprint and improve inference speed, only"},"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":"2507.11273","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-07-15T12:52:12Z","cross_cats_sorted":[],"title_canon_sha256":"a3d3ebda91651e066514a30808a7c91ce63a2856df44d7fac1ab7afacacc576c","abstract_canon_sha256":"b838fa756c2dd1211fd93b45095f0741168ed9f3c88bcf4a24dda9c95bc0be35"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:37:36.041259Z","signature_b64":"w/4THfcgD1hcMXw57GHl5DNVN6yAj2GQxS+P/feydMhvIjPqWN99qKpXOtwxCJw+XGdafJiRruMYyXmfcQ5tCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"668c2567346e656c9bf95d059718e4f22647b0e783a9fd2b294e477097b76b87","last_reissued_at":"2026-07-05T11:37:36.040727Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:37:36.040727Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"KV-Latent: Dimensional-level KV Cache Reduction with Frequency-aware Rotary Positional Embedding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Baoyuan Qi, Guoming Liu, Hai Zhao, Lefei Zhang, Luohe Shi, Zuchao Li","submitted_at":"2025-07-15T12:52:12Z","abstract_excerpt":"Large language models (LLMs) based on Transformer Decoders have become the preferred choice for conversational generative AI. Despite the overall superiority of the Decoder architecture, the gradually increasing Key-Value (KV) cache during inference has emerged as a primary efficiency bottleneck, both in aspects of memory consumption and data transfer bandwidth limitations. To address these challenges, we propose a paradigm called KV-Latent. By down-sampling the Key-Value vector dimensions into a latent space, we can significantly reduce the KV Cache footprint and improve inference speed, only"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.11273","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/2507.11273/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":"2507.11273","created_at":"2026-07-05T11:37:36.040794+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.11273v1","created_at":"2026-07-05T11:37:36.040794+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.11273","created_at":"2026-07-05T11:37:36.040794+00:00"},{"alias_kind":"pith_short_12","alias_value":"M2GCKZZUNZSW","created_at":"2026-07-05T11:37:36.040794+00:00"},{"alias_kind":"pith_short_16","alias_value":"M2GCKZZUNZSWZG7Z","created_at":"2026-07-05T11:37:36.040794+00:00"},{"alias_kind":"pith_short_8","alias_value":"M2GCKZZU","created_at":"2026-07-05T11:37:36.040794+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/M2GCKZZUNZSWZG7ZLUCZOGHE6I","json":"https://pith.science/pith/M2GCKZZUNZSWZG7ZLUCZOGHE6I.json","graph_json":"https://pith.science/api/pith-number/M2GCKZZUNZSWZG7ZLUCZOGHE6I/graph.json","events_json":"https://pith.science/api/pith-number/M2GCKZZUNZSWZG7ZLUCZOGHE6I/events.json","paper":"https://pith.science/paper/M2GCKZZU"},"agent_actions":{"view_html":"https://pith.science/pith/M2GCKZZUNZSWZG7ZLUCZOGHE6I","download_json":"https://pith.science/pith/M2GCKZZUNZSWZG7ZLUCZOGHE6I.json","view_paper":"https://pith.science/paper/M2GCKZZU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.11273&json=true","fetch_graph":"https://pith.science/api/pith-number/M2GCKZZUNZSWZG7ZLUCZOGHE6I/graph.json","fetch_events":"https://pith.science/api/pith-number/M2GCKZZUNZSWZG7ZLUCZOGHE6I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/M2GCKZZUNZSWZG7ZLUCZOGHE6I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/M2GCKZZUNZSWZG7ZLUCZOGHE6I/action/storage_attestation","attest_author":"https://pith.science/pith/M2GCKZZUNZSWZG7ZLUCZOGHE6I/action/author_attestation","sign_citation":"https://pith.science/pith/M2GCKZZUNZSWZG7ZLUCZOGHE6I/action/citation_signature","submit_replication":"https://pith.science/pith/M2GCKZZUNZSWZG7ZLUCZOGHE6I/action/replication_record"}},"created_at":"2026-07-05T11:37:36.040794+00:00","updated_at":"2026-07-05T11:37:36.040794+00:00"}