{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:H35TVOAN7TIHAI7EEBOS7STMPD","short_pith_number":"pith:H35TVOAN","schema_version":"1.0","canonical_sha256":"3efb3ab80dfcd07023e4205d2fca6c78e1e5debead92470aa0bacf45f79adb28","source":{"kind":"arxiv","id":"2507.07400","version":1},"attestation_state":"computed","paper":{"title":"KVFlow: Efficient Prefix Caching for Accelerating LLM-Based Multi-Agent Workflows","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.MA"],"primary_cat":"cs.DC","authors_text":"Ajjkumar Patel, Lianhui Qin, Wan-Lu Li, Yida Wang, Yipeng Shen, Yue Guan, Yufei Ding, Zaifeng Pan, Zhengding Hu","submitted_at":"2025-07-10T03:39:23Z","abstract_excerpt":"Large language model (LLM) based agentic workflows have become a popular paradigm for coordinating multiple specialized agents to solve complex tasks. To improve serving efficiency, existing LLM systems employ prefix caching to reuse key-value (KV) tensors corresponding to agents' fixed prompts, thereby avoiding redundant computation across repeated invocations. However, current systems typically evict KV caches using a Least Recently Used (LRU) policy, which fails to anticipate future agent usage and often discards KV caches shortly before their reuse. This leads to frequent cache misses and "},"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":"2507.07400","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2025-07-10T03:39:23Z","cross_cats_sorted":["cs.MA"],"title_canon_sha256":"6692c5a8253ab147cb6d768305770fabe0c3f069d59f7651cbaf6f7d04482f1d","abstract_canon_sha256":"338390cd8fb9527d9a0c5a5371ccab03cd212f8e4fdc8f45a90fb9ff8d1992a6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:34:57.220150Z","signature_b64":"uOYkREgS4a7Ktzq9Aq40xRSgCM5WN6a90CMI6EFTFFvKMX+ZF7Bck34qYZFy/Fo4X8xNcI1mKiMkkGiITwg6BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3efb3ab80dfcd07023e4205d2fca6c78e1e5debead92470aa0bacf45f79adb28","last_reissued_at":"2026-07-05T11:34:57.219681Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:34:57.219681Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"KVFlow: Efficient Prefix Caching for Accelerating LLM-Based Multi-Agent Workflows","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.MA"],"primary_cat":"cs.DC","authors_text":"Ajjkumar Patel, Lianhui Qin, Wan-Lu Li, Yida Wang, Yipeng Shen, Yue Guan, Yufei Ding, Zaifeng Pan, Zhengding Hu","submitted_at":"2025-07-10T03:39:23Z","abstract_excerpt":"Large language model (LLM) based agentic workflows have become a popular paradigm for coordinating multiple specialized agents to solve complex tasks. To improve serving efficiency, existing LLM systems employ prefix caching to reuse key-value (KV) tensors corresponding to agents' fixed prompts, thereby avoiding redundant computation across repeated invocations. However, current systems typically evict KV caches using a Least Recently Used (LRU) policy, which fails to anticipate future agent usage and often discards KV caches shortly before their reuse. This leads to frequent cache misses and "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.07400","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/2507.07400/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":"2507.07400","created_at":"2026-07-05T11:34:57.219738+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.07400v1","created_at":"2026-07-05T11:34:57.219738+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.07400","created_at":"2026-07-05T11:34:57.219738+00:00"},{"alias_kind":"pith_short_12","alias_value":"H35TVOAN7TIH","created_at":"2026-07-05T11:34:57.219738+00:00"},{"alias_kind":"pith_short_16","alias_value":"H35TVOAN7TIHAI7E","created_at":"2026-07-05T11:34:57.219738+00:00"},{"alias_kind":"pith_short_8","alias_value":"H35TVOAN","created_at":"2026-07-05T11:34:57.219738+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":11,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.00151","citing_title":"SmoothAgent: Efficient Long-Horizon LLM-Based Agent Serving with Lookahead Context Engineering","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05158","citing_title":"Streaming Communication in Multi-Agent Reasoning","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24259","citing_title":"Resident KV Claims: A Conformance Contract for Future Reuse under Active KV Pressure","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29640","citing_title":"VikingMem: A Memory Base Management System for Stateful LLM-based Applications","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18825","citing_title":"Not All Tokens Are Worth Caching: Learning Semantic-Aware Eviction for LLM Prefix Caches","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17613","citing_title":"VeriCache: Turning Lossy KV Cache into Lossless LLM Inference","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08581","citing_title":"PRISM: Fast Online LLM Serving via Scheduling-Memory Co-design","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26103","citing_title":"AMMA: A Multi-Chiplet Memory-Centric Architecture for Low-Latency 1M Context Attention Serving","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24971","citing_title":"PolyKV: A Shared Asymmetrically-Compressed KV Cache Pool for Multi-Agent LLM Inference","ref_index":7,"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":47,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17353","citing_title":"Hive: A Multi-Agent Infrastructure for Algorithm- and Task-Level Scaling","ref_index":32,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/H35TVOAN7TIHAI7EEBOS7STMPD","json":"https://pith.science/pith/H35TVOAN7TIHAI7EEBOS7STMPD.json","graph_json":"https://pith.science/api/pith-number/H35TVOAN7TIHAI7EEBOS7STMPD/graph.json","events_json":"https://pith.science/api/pith-number/H35TVOAN7TIHAI7EEBOS7STMPD/events.json","paper":"https://pith.science/paper/H35TVOAN"},"agent_actions":{"view_html":"https://pith.science/pith/H35TVOAN7TIHAI7EEBOS7STMPD","download_json":"https://pith.science/pith/H35TVOAN7TIHAI7EEBOS7STMPD.json","view_paper":"https://pith.science/paper/H35TVOAN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.07400&json=true","fetch_graph":"https://pith.science/api/pith-number/H35TVOAN7TIHAI7EEBOS7STMPD/graph.json","fetch_events":"https://pith.science/api/pith-number/H35TVOAN7TIHAI7EEBOS7STMPD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/H35TVOAN7TIHAI7EEBOS7STMPD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/H35TVOAN7TIHAI7EEBOS7STMPD/action/storage_attestation","attest_author":"https://pith.science/pith/H35TVOAN7TIHAI7EEBOS7STMPD/action/author_attestation","sign_citation":"https://pith.science/pith/H35TVOAN7TIHAI7EEBOS7STMPD/action/citation_signature","submit_replication":"https://pith.science/pith/H35TVOAN7TIHAI7EEBOS7STMPD/action/replication_record"}},"created_at":"2026-07-05T11:34:57.219738+00:00","updated_at":"2026-07-05T11:34:57.219738+00:00"}