{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4JIKGC4NWX2TSZM7QRNYUWUQUV","short_pith_number":"pith:4JIKGC4N","schema_version":"1.0","canonical_sha256":"e250a30b8db5f539659f845b8a5a90a575a148cb59345e42130f317fbb2c2198","source":{"kind":"arxiv","id":"2502.08773","version":2},"attestation_state":"computed","paper":{"title":"Universal Model Routing for Efficient LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Aditya Krishna Menon, Alec Go, Ankit Singh Rawat, Chen-Yu Lee, Congchao Wang, Harikrishna Narasimhan, Jeevesh Juneja, Pradeep Shenoy, Rina Panigrahy, Sanjiv Kumar, Wittawat Jitkrittum, Zifeng Wang","submitted_at":"2025-02-12T20:30:28Z","abstract_excerpt":"Model routing is a simple technique for reducing the inference cost of large language models (LLMs), wherein one maintains a pool of candidate LLMs, and learns to route each prompt to the smallest feasible LLM. Existing works focus on learning a router for a fixed pool of LLMs. In this paper, we consider the problem of dynamic routing, where new, previously unobserved LLMs are available at test time. We propose UniRoute, a new approach to this problem that relies on representing each LLM as a feature vector, derived based on predictions on a set of representative prompts. Based on this, we det"},"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":"2502.08773","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-12T20:30:28Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"b485ff06966b09029838f70a41ea2b53809ec3ef0e8c125f95cc40c8d8a38648","abstract_canon_sha256":"6243323ea3b1db6c1bf88dede7488239d97dd8de96d9998a17178e87c654c939"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:41:21.394523Z","signature_b64":"luyrCNAOnNhWIdHrO0rXioEEo7HQ/AHnyiwUQvbEE1bgEQI/VMk/qyo4EHo7u6Ym9FK9adUosfjg/OhlropVCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e250a30b8db5f539659f845b8a5a90a575a148cb59345e42130f317fbb2c2198","last_reissued_at":"2026-07-05T11:41:21.393976Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:41:21.393976Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Universal Model Routing for Efficient LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Aditya Krishna Menon, Alec Go, Ankit Singh Rawat, Chen-Yu Lee, Congchao Wang, Harikrishna Narasimhan, Jeevesh Juneja, Pradeep Shenoy, Rina Panigrahy, Sanjiv Kumar, Wittawat Jitkrittum, Zifeng Wang","submitted_at":"2025-02-12T20:30:28Z","abstract_excerpt":"Model routing is a simple technique for reducing the inference cost of large language models (LLMs), wherein one maintains a pool of candidate LLMs, and learns to route each prompt to the smallest feasible LLM. Existing works focus on learning a router for a fixed pool of LLMs. In this paper, we consider the problem of dynamic routing, where new, previously unobserved LLMs are available at test time. We propose UniRoute, a new approach to this problem that relies on representing each LLM as a feature vector, derived based on predictions on a set of representative prompts. Based on this, we det"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.08773","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/2502.08773/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":"2502.08773","created_at":"2026-07-05T11:41:21.394043+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.08773v2","created_at":"2026-07-05T11:41:21.394043+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.08773","created_at":"2026-07-05T11:41:21.394043+00:00"},{"alias_kind":"pith_short_12","alias_value":"4JIKGC4NWX2T","created_at":"2026-07-05T11:41:21.394043+00:00"},{"alias_kind":"pith_short_16","alias_value":"4JIKGC4NWX2TSZM7","created_at":"2026-07-05T11:41:21.394043+00:00"},{"alias_kind":"pith_short_8","alias_value":"4JIKGC4N","created_at":"2026-07-05T11:41:21.394043+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":15,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06145","citing_title":"Prompting Complexity: Shortest Prompts for Texts and Behaviors in LLMs","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2606.26836","citing_title":"The Capability Frontier: Benchmarks Miss 82% of Model Performance","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.17949","citing_title":"RouteBalance: Fused Model Routing and Load Balancing for Heterogeneous LLM Serving","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00053","citing_title":"SWE-Router: Routing in Multi-turn Agentic Software Engineering Tasks","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07711","citing_title":"Rosetta Memory: Adaptive Memory for Cross-LLM Agents","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06098","citing_title":"IR3DE: A Linear Router for Large Language Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07587","citing_title":"The Routing Plateau: Understanding and Breaking the Accuracy Limits of LLM Routers","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2507.14200","citing_title":"A Scalable Multi-LLM Collaboration System with Retrieval-based Selection and Exploration-Exploitation-Driven Enhancement","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14241","citing_title":"Latency-Quality Routing for Functionally Equivalent Tools in LLM Agents","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07112","citing_title":"Switchcraft: AI Model Router for Agentic Tool Calling","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07180","citing_title":"Learning Agent Routing From Early Experience","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07171","citing_title":"Cost-Ordered Feasibility for Multi-Armed Bandits with Cost Subsidy","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07805","citing_title":"Flexible Routing via Uncertainty Decomposition","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15499","citing_title":"SecureRouter: Encrypted Routing for Efficient Secure Inference","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15728","citing_title":"Privacy-Preserving LLMs Routing","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4JIKGC4NWX2TSZM7QRNYUWUQUV","json":"https://pith.science/pith/4JIKGC4NWX2TSZM7QRNYUWUQUV.json","graph_json":"https://pith.science/api/pith-number/4JIKGC4NWX2TSZM7QRNYUWUQUV/graph.json","events_json":"https://pith.science/api/pith-number/4JIKGC4NWX2TSZM7QRNYUWUQUV/events.json","paper":"https://pith.science/paper/4JIKGC4N"},"agent_actions":{"view_html":"https://pith.science/pith/4JIKGC4NWX2TSZM7QRNYUWUQUV","download_json":"https://pith.science/pith/4JIKGC4NWX2TSZM7QRNYUWUQUV.json","view_paper":"https://pith.science/paper/4JIKGC4N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.08773&json=true","fetch_graph":"https://pith.science/api/pith-number/4JIKGC4NWX2TSZM7QRNYUWUQUV/graph.json","fetch_events":"https://pith.science/api/pith-number/4JIKGC4NWX2TSZM7QRNYUWUQUV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4JIKGC4NWX2TSZM7QRNYUWUQUV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4JIKGC4NWX2TSZM7QRNYUWUQUV/action/storage_attestation","attest_author":"https://pith.science/pith/4JIKGC4NWX2TSZM7QRNYUWUQUV/action/author_attestation","sign_citation":"https://pith.science/pith/4JIKGC4NWX2TSZM7QRNYUWUQUV/action/citation_signature","submit_replication":"https://pith.science/pith/4JIKGC4NWX2TSZM7QRNYUWUQUV/action/replication_record"}},"created_at":"2026-07-05T11:41:21.394043+00:00","updated_at":"2026-07-05T11:41:21.394043+00:00"}