{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:F22MPUSCVMTNP2ZWQIPGUBRPBC","short_pith_number":"pith:F22MPUSC","schema_version":"1.0","canonical_sha256":"2eb4c7d242ab26d7eb36821e6a062f089a330e595e552e25bc184ccc03b216d8","source":{"kind":"arxiv","id":"2505.03756","version":1},"attestation_state":"computed","paper":{"title":"Improving the Serving Performance of Multi-LoRA Large Language Models via Efficient LoRA and KV Cache Management","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.PF"],"primary_cat":"cs.AR","authors_text":"Hang Zhang, Jiuchen Shi, Minyi Guo, Quan Chen, Yixiao Wang, Yizhou Shan","submitted_at":"2025-04-19T13:17:34Z","abstract_excerpt":"Multiple Low-Rank Adapters (Multi-LoRAs) are gaining popularity for task-specific Large Language Model (LLM) applications. For multi-LoRA serving, caching hot KV caches and LoRA adapters in high bandwidth memory of accelerations can improve inference performance. However, existing Multi-LoRA inference systems fail to optimize serving performance like Time-To-First-Toke (TTFT), neglecting usage dependencies when caching LoRAs and KVs. We therefore propose FASTLIBRA, a Multi-LoRA caching system to optimize the serving performance. FASTLIBRA comprises a dependency-aware cache manager and a perfor"},"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":"2505.03756","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AR","submitted_at":"2025-04-19T13:17:34Z","cross_cats_sorted":["cs.AI","cs.LG","cs.PF"],"title_canon_sha256":"bda410aaa4f6721f12ca9933c6ac632499c61955f9f72d5ad675c86904246b4c","abstract_canon_sha256":"b23c49586ee5e71c0df0ac911bedd1b3b653b4055b61101a22d55f4fb9b13984"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:59:30.460504Z","signature_b64":"uOTlEPVksTGjuiBnJPi5DqHsrrSTlNE2YTPTf4NDNqy1WeiQeD1S3WyohzPDSxJscAb1CCeO/SN1Q/ln7KJvDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2eb4c7d242ab26d7eb36821e6a062f089a330e595e552e25bc184ccc03b216d8","last_reissued_at":"2026-07-05T10:59:30.460009Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:59:30.460009Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improving the Serving Performance of Multi-LoRA Large Language Models via Efficient LoRA and KV Cache Management","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.PF"],"primary_cat":"cs.AR","authors_text":"Hang Zhang, Jiuchen Shi, Minyi Guo, Quan Chen, Yixiao Wang, Yizhou Shan","submitted_at":"2025-04-19T13:17:34Z","abstract_excerpt":"Multiple Low-Rank Adapters (Multi-LoRAs) are gaining popularity for task-specific Large Language Model (LLM) applications. For multi-LoRA serving, caching hot KV caches and LoRA adapters in high bandwidth memory of accelerations can improve inference performance. However, existing Multi-LoRA inference systems fail to optimize serving performance like Time-To-First-Toke (TTFT), neglecting usage dependencies when caching LoRAs and KVs. We therefore propose FASTLIBRA, a Multi-LoRA caching system to optimize the serving performance. FASTLIBRA comprises a dependency-aware cache manager and a perfor"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.03756","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/2505.03756/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":"2505.03756","created_at":"2026-07-05T10:59:30.460066+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.03756v1","created_at":"2026-07-05T10:59:30.460066+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.03756","created_at":"2026-07-05T10:59:30.460066+00:00"},{"alias_kind":"pith_short_12","alias_value":"F22MPUSCVMTN","created_at":"2026-07-05T10:59:30.460066+00:00"},{"alias_kind":"pith_short_16","alias_value":"F22MPUSCVMTNP2ZW","created_at":"2026-07-05T10:59:30.460066+00:00"},{"alias_kind":"pith_short_8","alias_value":"F22MPUSC","created_at":"2026-07-05T10:59:30.460066+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.13779","citing_title":"MinT: Managed Infrastructure for Training and Serving Millions of LLMs","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14217","citing_title":"PreFT: Prefill-only finetuning for efficient inference","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13779","citing_title":"MinT: Managed Infrastructure for Training and Serving Millions of LLMs","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07173","citing_title":"InfiniLoRA: Disaggregated Multi-LoRA Serving for Large Language Models","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06370","citing_title":"ForkKV: Scaling Multi-LoRA Agent Serving via Copy-on-Write Disaggregated KV Cache","ref_index":76,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16583","citing_title":"POLAR: Online Learning for LoRA Adapter Caching and Routing in Edge LLM Serving","ref_index":29,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/F22MPUSCVMTNP2ZWQIPGUBRPBC","json":"https://pith.science/pith/F22MPUSCVMTNP2ZWQIPGUBRPBC.json","graph_json":"https://pith.science/api/pith-number/F22MPUSCVMTNP2ZWQIPGUBRPBC/graph.json","events_json":"https://pith.science/api/pith-number/F22MPUSCVMTNP2ZWQIPGUBRPBC/events.json","paper":"https://pith.science/paper/F22MPUSC"},"agent_actions":{"view_html":"https://pith.science/pith/F22MPUSCVMTNP2ZWQIPGUBRPBC","download_json":"https://pith.science/pith/F22MPUSCVMTNP2ZWQIPGUBRPBC.json","view_paper":"https://pith.science/paper/F22MPUSC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.03756&json=true","fetch_graph":"https://pith.science/api/pith-number/F22MPUSCVMTNP2ZWQIPGUBRPBC/graph.json","fetch_events":"https://pith.science/api/pith-number/F22MPUSCVMTNP2ZWQIPGUBRPBC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F22MPUSCVMTNP2ZWQIPGUBRPBC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F22MPUSCVMTNP2ZWQIPGUBRPBC/action/storage_attestation","attest_author":"https://pith.science/pith/F22MPUSCVMTNP2ZWQIPGUBRPBC/action/author_attestation","sign_citation":"https://pith.science/pith/F22MPUSCVMTNP2ZWQIPGUBRPBC/action/citation_signature","submit_replication":"https://pith.science/pith/F22MPUSCVMTNP2ZWQIPGUBRPBC/action/replication_record"}},"created_at":"2026-07-05T10:59:30.460066+00:00","updated_at":"2026-07-05T10:59:30.460066+00:00"}