{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:IY7VYG573A2MVPXJCG2J7VIZXD","short_pith_number":"pith:IY7VYG57","schema_version":"1.0","canonical_sha256":"463f5c1bbfd834cabee911b49fd519b8ef55271687f74bd7386ee83d53733bbe","source":{"kind":"arxiv","id":"2602.01053","version":2},"attestation_state":"computed","paper":{"title":"LRAgent: Efficient KV Cache Sharing for Multi-LoRA LLM Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hyeongju Ha, Hyesung Jeon, Jae-Joon Kim","submitted_at":"2026-02-01T06:36:46Z","abstract_excerpt":"Role specialization in multi-LLM agent systems is often realized via multi-LoRA, where agents share a pretrained backbone and differ only by lightweight adapters. Despite sharing base model weights, each agent independently builds and stores its own KV cache for the same long, tool-augmented trajectories, incurring substantial memory and compute overhead. Existing KV cache sharing methods largely overlook this multi-LoRA setting. We observe that, cache differences across agents are dominated by adapter outputs, while activations from the shared pretrained backbone remain highly similar. Based "},"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":"2602.01053","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-01T06:36:46Z","cross_cats_sorted":[],"title_canon_sha256":"0a2f2d9944372810dfea3816593a946bfde73017d21363a94c8b193edac372e2","abstract_canon_sha256":"8ddc7c0ecc02ce5c2dab727e362ecc571e1ff8c5260ee82e8e0eb3753ed3f306"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:04:15.568733Z","signature_b64":"Y1j1jNf7/cP/LT3ueYh6tCdgDZ/LLbtuVerZnfcqwgJEndYRynYQsHgRxPeNoS3Qo+FbF1/tj4SI6BLMtS1EAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"463f5c1bbfd834cabee911b49fd519b8ef55271687f74bd7386ee83d53733bbe","last_reissued_at":"2026-06-02T01:04:15.567883Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:04:15.567883Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LRAgent: Efficient KV Cache Sharing for Multi-LoRA LLM Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hyeongju Ha, Hyesung Jeon, Jae-Joon Kim","submitted_at":"2026-02-01T06:36:46Z","abstract_excerpt":"Role specialization in multi-LLM agent systems is often realized via multi-LoRA, where agents share a pretrained backbone and differ only by lightweight adapters. Despite sharing base model weights, each agent independently builds and stores its own KV cache for the same long, tool-augmented trajectories, incurring substantial memory and compute overhead. Existing KV cache sharing methods largely overlook this multi-LoRA setting. We observe that, cache differences across agents are dominated by adapter outputs, while activations from the shared pretrained backbone remain highly similar. Based "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.01053","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/2602.01053/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":"2602.01053","created_at":"2026-06-02T01:04:15.568210+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.01053v2","created_at":"2026-06-02T01:04:15.568210+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.01053","created_at":"2026-06-02T01:04:15.568210+00:00"},{"alias_kind":"pith_short_12","alias_value":"IY7VYG573A2M","created_at":"2026-06-02T01:04:15.568210+00:00"},{"alias_kind":"pith_short_16","alias_value":"IY7VYG573A2MVPXJ","created_at":"2026-06-02T01:04:15.568210+00:00"},{"alias_kind":"pith_short_8","alias_value":"IY7VYG57","created_at":"2026-06-02T01:04:15.568210+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"2604.03143","citing_title":"TokenDance: Scaling Multi-Agent LLM Serving via Collective KV Cache Sharing","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09104","citing_title":"Token Economics for LLM Agents: A Dual-View Study from Computing and Economics","ref_index":114,"is_internal_anchor":true},{"citing_arxiv_id":"2605.03884","citing_title":"QKVShare: Quantized KV-Cache Handoff for Multi-Agent On-Device LLMs","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2604.06370","citing_title":"ForkKV: Scaling Multi-LoRA Agent Serving via Copy-on-Write Disaggregated KV Cache","ref_index":23,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IY7VYG573A2MVPXJCG2J7VIZXD","json":"https://pith.science/pith/IY7VYG573A2MVPXJCG2J7VIZXD.json","graph_json":"https://pith.science/api/pith-number/IY7VYG573A2MVPXJCG2J7VIZXD/graph.json","events_json":"https://pith.science/api/pith-number/IY7VYG573A2MVPXJCG2J7VIZXD/events.json","paper":"https://pith.science/paper/IY7VYG57"},"agent_actions":{"view_html":"https://pith.science/pith/IY7VYG573A2MVPXJCG2J7VIZXD","download_json":"https://pith.science/pith/IY7VYG573A2MVPXJCG2J7VIZXD.json","view_paper":"https://pith.science/paper/IY7VYG57","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.01053&json=true","fetch_graph":"https://pith.science/api/pith-number/IY7VYG573A2MVPXJCG2J7VIZXD/graph.json","fetch_events":"https://pith.science/api/pith-number/IY7VYG573A2MVPXJCG2J7VIZXD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IY7VYG573A2MVPXJCG2J7VIZXD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IY7VYG573A2MVPXJCG2J7VIZXD/action/storage_attestation","attest_author":"https://pith.science/pith/IY7VYG573A2MVPXJCG2J7VIZXD/action/author_attestation","sign_citation":"https://pith.science/pith/IY7VYG573A2MVPXJCG2J7VIZXD/action/citation_signature","submit_replication":"https://pith.science/pith/IY7VYG573A2MVPXJCG2J7VIZXD/action/replication_record"}},"created_at":"2026-06-02T01:04:15.568210+00:00","updated_at":"2026-06-02T01:04:15.568210+00:00"}