{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LRYBKRX23MSSRLZIMDLMB422Z5","short_pith_number":"pith:LRYBKRX2","schema_version":"1.0","canonical_sha256":"5c701546fadb2528af2860d6c0f35acf635655fff6713bb248ea2144fe6c4902","source":{"kind":"arxiv","id":"2509.09782","version":1},"attestation_state":"computed","paper":{"title":"One Head, Many Models: Cross-Attention Routing for Cost-Aware LLM Selection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aditya Mate, Ehi Nosakhare, Keerthy Kaushik Dasoju, Mani Kishan Ghantasala, Niti Mangwani, Roshini Pulishetty, Sadid Hasan, Somya Chatterjee, Soundar Srinivasan, Vishal Garimella, Yue Kang","submitted_at":"2025-09-11T18:29:09Z","abstract_excerpt":"The proliferation of large language models (LLMs) with varying computational costs and performance profiles presents a critical challenge for scalable, cost-effective deployment in real-world applications. We introduce a unified routing framework that leverages a single-head cross-attention mechanism to jointly model query and model embeddings, enabling dynamic selection of the optimal LLM for each input query. Our approach is evaluated on RouterBench, a large-scale, publicly available benchmark encompassing diverse LLM pools and domains. By explicitly capturing fine-grained query-model intera"},"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":"2509.09782","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-09-11T18:29:09Z","cross_cats_sorted":[],"title_canon_sha256":"e6df045a24af73f0830c257cf71362c75942e4e4f285d3df9172633421c9d98c","abstract_canon_sha256":"dce9df079e98b6941326216f23da1f3371935c17dda6f52a9e936d34981ef27b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:09:49.216586Z","signature_b64":"tSn3VjvVjIsg7JUD+QumwzpmLsxBDTslqqrhwWvrdRTcYa0xQ7RQ8yq2pXpioc7sCypAbZPwDMlZWaAKruFOCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5c701546fadb2528af2860d6c0f35acf635655fff6713bb248ea2144fe6c4902","last_reissued_at":"2026-07-05T12:09:49.216029Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:09:49.216029Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"One Head, Many Models: Cross-Attention Routing for Cost-Aware LLM Selection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aditya Mate, Ehi Nosakhare, Keerthy Kaushik Dasoju, Mani Kishan Ghantasala, Niti Mangwani, Roshini Pulishetty, Sadid Hasan, Somya Chatterjee, Soundar Srinivasan, Vishal Garimella, Yue Kang","submitted_at":"2025-09-11T18:29:09Z","abstract_excerpt":"The proliferation of large language models (LLMs) with varying computational costs and performance profiles presents a critical challenge for scalable, cost-effective deployment in real-world applications. We introduce a unified routing framework that leverages a single-head cross-attention mechanism to jointly model query and model embeddings, enabling dynamic selection of the optimal LLM for each input query. Our approach is evaluated on RouterBench, a large-scale, publicly available benchmark encompassing diverse LLM pools and domains. By explicitly capturing fine-grained query-model intera"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.09782","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/2509.09782/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":"2509.09782","created_at":"2026-07-05T12:09:49.216100+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.09782v1","created_at":"2026-07-05T12:09:49.216100+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.09782","created_at":"2026-07-05T12:09:49.216100+00:00"},{"alias_kind":"pith_short_12","alias_value":"LRYBKRX23MSS","created_at":"2026-07-05T12:09:49.216100+00:00"},{"alias_kind":"pith_short_16","alias_value":"LRYBKRX23MSSRLZI","created_at":"2026-07-05T12:09:49.216100+00:00"},{"alias_kind":"pith_short_8","alias_value":"LRYBKRX2","created_at":"2026-07-05T12:09:49.216100+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.26955","citing_title":"Policy-Governed LLM Routing with Intent Matching for Instrument Laboratories","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LRYBKRX23MSSRLZIMDLMB422Z5","json":"https://pith.science/pith/LRYBKRX23MSSRLZIMDLMB422Z5.json","graph_json":"https://pith.science/api/pith-number/LRYBKRX23MSSRLZIMDLMB422Z5/graph.json","events_json":"https://pith.science/api/pith-number/LRYBKRX23MSSRLZIMDLMB422Z5/events.json","paper":"https://pith.science/paper/LRYBKRX2"},"agent_actions":{"view_html":"https://pith.science/pith/LRYBKRX23MSSRLZIMDLMB422Z5","download_json":"https://pith.science/pith/LRYBKRX23MSSRLZIMDLMB422Z5.json","view_paper":"https://pith.science/paper/LRYBKRX2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.09782&json=true","fetch_graph":"https://pith.science/api/pith-number/LRYBKRX23MSSRLZIMDLMB422Z5/graph.json","fetch_events":"https://pith.science/api/pith-number/LRYBKRX23MSSRLZIMDLMB422Z5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LRYBKRX23MSSRLZIMDLMB422Z5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LRYBKRX23MSSRLZIMDLMB422Z5/action/storage_attestation","attest_author":"https://pith.science/pith/LRYBKRX23MSSRLZIMDLMB422Z5/action/author_attestation","sign_citation":"https://pith.science/pith/LRYBKRX23MSSRLZIMDLMB422Z5/action/citation_signature","submit_replication":"https://pith.science/pith/LRYBKRX23MSSRLZIMDLMB422Z5/action/replication_record"}},"created_at":"2026-07-05T12:09:49.216100+00:00","updated_at":"2026-07-05T12:09:49.216100+00:00"}