{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5W4D6FR2F6EY7OEVDSZSOUFIY6","short_pith_number":"pith:5W4D6FR2","schema_version":"1.0","canonical_sha256":"edb83f163a2f898fb8951cb32750a8c7883345e0014b3a7cf316746b2304c14b","source":{"kind":"arxiv","id":"2504.07113","version":1},"attestation_state":"computed","paper":{"title":"How Robust Are Router-LLMs? Analysis of the Fragility of LLM Routing Capabilities","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DB"],"primary_cat":"cs.CL","authors_text":"Aly M. Kassem, Bernhard Sch\\\"olkopf, Zhijing Jin","submitted_at":"2025-03-20T19:52:30Z","abstract_excerpt":"Large language model (LLM) routing has emerged as a crucial strategy for balancing computational costs with performance by dynamically assigning queries to the most appropriate model based on query complexity. Despite recent advances showing that preference-data-based routers can outperform traditional methods, current evaluation benchmarks remain limited. They largely focus on general model capabilities while overlooking task-specific behaviors and critical concerns such as privacy, safety, and potential backdoor vulnerabilities introduced through preference data. In response, we propose the "},"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":"2504.07113","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-03-20T19:52:30Z","cross_cats_sorted":["cs.DB"],"title_canon_sha256":"7bbc66c47a0b6ffce0de916ce87e24e92980c416fd1a81e85821a2b7f760fd8f","abstract_canon_sha256":"3f2be22fe3fbdf63067f57c4aec711c6d3c9e233cce7bb2e563ade0ba4e66a55"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:47:03.492541Z","signature_b64":"sAiD+xx+300RxEHBxlqVEE9zLI/yePfuyMoJJdqxMDHuyN8RS/qR9nm5+1uv9qdrbixbkg4RiXgQh/5V4LLqDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"edb83f163a2f898fb8951cb32750a8c7883345e0014b3a7cf316746b2304c14b","last_reissued_at":"2026-07-05T10:47:03.491948Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:47:03.491948Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Robust Are Router-LLMs? Analysis of the Fragility of LLM Routing Capabilities","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DB"],"primary_cat":"cs.CL","authors_text":"Aly M. Kassem, Bernhard Sch\\\"olkopf, Zhijing Jin","submitted_at":"2025-03-20T19:52:30Z","abstract_excerpt":"Large language model (LLM) routing has emerged as a crucial strategy for balancing computational costs with performance by dynamically assigning queries to the most appropriate model based on query complexity. Despite recent advances showing that preference-data-based routers can outperform traditional methods, current evaluation benchmarks remain limited. They largely focus on general model capabilities while overlooking task-specific behaviors and critical concerns such as privacy, safety, and potential backdoor vulnerabilities introduced through preference data. In response, we propose the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.07113","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/2504.07113/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":"2504.07113","created_at":"2026-07-05T10:47:03.492033+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.07113v1","created_at":"2026-07-05T10:47:03.492033+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.07113","created_at":"2026-07-05T10:47:03.492033+00:00"},{"alias_kind":"pith_short_12","alias_value":"5W4D6FR2F6EY","created_at":"2026-07-05T10:47:03.492033+00:00"},{"alias_kind":"pith_short_16","alias_value":"5W4D6FR2F6EY7OEV","created_at":"2026-07-05T10:47:03.492033+00:00"},{"alias_kind":"pith_short_8","alias_value":"5W4D6FR2","created_at":"2026-07-05T10:47:03.492033+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.00053","citing_title":"SWE-Router: Routing in Multi-turn Agentic Software Engineering Tasks","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06924","citing_title":"From Sampled Outcomes to Capability Distributions: Rethinking Supervision for LLM Routing","ref_index":137,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15022","citing_title":"Route to Rome Attack: Directing LLM Routers to Expensive Models via Adversarial Suffix Optimization","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5W4D6FR2F6EY7OEVDSZSOUFIY6","json":"https://pith.science/pith/5W4D6FR2F6EY7OEVDSZSOUFIY6.json","graph_json":"https://pith.science/api/pith-number/5W4D6FR2F6EY7OEVDSZSOUFIY6/graph.json","events_json":"https://pith.science/api/pith-number/5W4D6FR2F6EY7OEVDSZSOUFIY6/events.json","paper":"https://pith.science/paper/5W4D6FR2"},"agent_actions":{"view_html":"https://pith.science/pith/5W4D6FR2F6EY7OEVDSZSOUFIY6","download_json":"https://pith.science/pith/5W4D6FR2F6EY7OEVDSZSOUFIY6.json","view_paper":"https://pith.science/paper/5W4D6FR2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.07113&json=true","fetch_graph":"https://pith.science/api/pith-number/5W4D6FR2F6EY7OEVDSZSOUFIY6/graph.json","fetch_events":"https://pith.science/api/pith-number/5W4D6FR2F6EY7OEVDSZSOUFIY6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5W4D6FR2F6EY7OEVDSZSOUFIY6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5W4D6FR2F6EY7OEVDSZSOUFIY6/action/storage_attestation","attest_author":"https://pith.science/pith/5W4D6FR2F6EY7OEVDSZSOUFIY6/action/author_attestation","sign_citation":"https://pith.science/pith/5W4D6FR2F6EY7OEVDSZSOUFIY6/action/citation_signature","submit_replication":"https://pith.science/pith/5W4D6FR2F6EY7OEVDSZSOUFIY6/action/replication_record"}},"created_at":"2026-07-05T10:47:03.492033+00:00","updated_at":"2026-07-05T10:47:03.492033+00:00"}