{"work":{"id":"f6b91d8a-ce78-4b5e-9dae-6dd8a22cc087","openalex_id":null,"doi":null,"arxiv_id":"2406.18665","raw_key":null,"title":"RouteLLM: Learning to Route LLMs with Preference Data","authors":null,"authors_text":"Isaac Ong, Amjad Almahairi, Vincent Wu, Wei-Lin Chiang, Tianhao Wu, Joseph E. Gonzalez","year":2024,"venue":"cs.LG","abstract":"Large language models (LLMs) exhibit impressive capabilities across a wide range of tasks, yet the choice of which model to use often involves a trade-off between performance and cost. More powerful models, though effective, come with higher expenses, while less capable models are more cost-effective. To address this dilemma, we propose several efficient router models that dynamically select between a stronger and a weaker LLM during inference, aiming to optimize the balance between cost and response quality. We develop a training framework for these routers leveraging human preference data and data augmentation techniques to enhance performance. Our evaluation on widely-recognized benchmarks shows that our approach significantly reduces costs-by over 2 times in certain cases-without compromising the quality of responses. Interestingly, our router models also demonstrate significant transfer learning capabilities, maintaining their performance even when the strong and weak models are changed at test time. This highlights the potential of these routers to provide a cost-effective yet high-performance solution for deploying LLMs.","external_url":"https://arxiv.org/abs/2406.18665","cited_by_count":null,"metadata_source":"pith","metadata_fetched_at":"2026-05-25T04:30:19.988379+00:00","pith_arxiv_id":"2406.18665","created_at":"2026-05-09T06:15:38.426678+00:00","updated_at":"2026-05-25T04:30:19.988379+00:00","title_quality_ok":true,"display_title":"RouteLLM: Learning to Route LLMs with Preference Data","render_title":"RouteLLM: Learning to Route LLMs with Preference Data"},"hub":{"state":{"work_id":"f6b91d8a-ce78-4b5e-9dae-6dd8a22cc087","tier":"hub","tier_reason":"10+ Pith inbound or 1,000+ external citations","pith_inbound_count":49,"external_cited_by_count":null,"distinct_field_count":12,"first_pith_cited_at":"2024-07-31T17:57:25+00:00","last_pith_cited_at":"2026-05-22T06:47:19+00:00","author_build_status":"not_needed","summary_status":"needed","contexts_status":"needed","graph_status":"needed","ask_index_status":"not_needed","reader_status":"not_needed","recognition_status":"not_needed","updated_at":"2026-05-29T22:50:29.720399+00:00","tier_text":"hub"},"tier":"hub","role_counts":[{"context_role":"background","n":8},{"context_role":"baseline","n":2}],"polarity_counts":[{"context_polarity":"background","n":8},{"context_polarity":"baseline","n":2}],"runs":{"context_extract":{"job_type":"context_extract","status":"succeeded","result":{"enqueued_papers":25},"error":null,"updated_at":"2026-05-14T18:29:59.938749+00:00"},"graph_features":{"job_type":"graph_features","status":"succeeded","result":{"co_cited":[{"title":"FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance","work_id":"5c567414-35de-4373-ae7a-1790da594e8d","shared_citers":18},{"title":"Huang, L., Yu, W., Ma, W., Zhong, W., Feng, Z., Wang, H., Chen, Q., Peng, W., Feng, X., Qin, B., and Liu, T","work_id":"ab0c1f58-f46a-4885-8afb-6a26e45608fb","shared_citers":8},{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":8},{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":7},{"title":"Hybrid llm: Cost-efficient and quality- aware query routing","work_id":"9dda325d-c570-4795-a95c-95c467dd09b4","shared_citers":7},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":6},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":6},{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":6},{"title":"AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation","work_id":"92b7eb9c-c3d8-4518-a376-06fa15dd895b","shared_citers":5},{"title":"Scaling Laws for Neural Language Models","work_id":"b7dd8749-9c45-4977-ab9b-64478dce1ae8","shared_citers":5},{"title":"Automix: Automatically mixing language models","work_id":"721159b7-a5b2-42e0-94e9-8aa5e1088d57","shared_citers":4},{"title":"MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework","work_id":"891b9780-a800-4e3c-bba0-53597ab8dc98","shared_citers":4},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":4},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":3},{"title":"Efficiently serving llm reasoning programs with certaindex","work_id":"57494d13-0d57-400a-8e38-67a61f1677d1","shared_citers":3},{"title":"Gemma: Open Models Based on Gemini Research and Technology","work_id":"a9ea2870-df28-40b8-a9e0-a7e9a116f793","shared_citers":3},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":3},{"title":"Graphrouter: A graph-based router for llm selections","work_id":"ed7f55c7-58fe-463b-af50-633c54486e8e","shared_citers":3},{"title":"Kimi k1.5: Scaling Reinforcement Learning with LLMs","work_id":"bff96ab1-bd6a-4585-be23-74fdb51969c7","shared_citers":3},{"title":"KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache","work_id":"735737c3-24e5-41c3-ab4f-04edcb36731c","shared_citers":3},{"title":"Pan, H., Tennenholtz, G., Mannor, S., Chi, C.-W., Brekel- mans, R., Shah, P., and Tewari, A","work_id":"86cd33f5-f0be-4d28-b775-10e356b31e74","shared_citers":3},{"title":"Qwen2.5 Technical Report","work_id":"d8432992-4980-4a81-85c7-9fa2c2b87f85","shared_citers":3},{"title":"ReAct: Synergizing Reasoning and Acting in Language Models","work_id":"407a2351-25f1-497d-b611-f77d0292a8e6","shared_citers":3},{"title":"Reward-guided speculative decoding for efficient llm reasoning","work_id":"460cb265-1bb5-4de5-a476-57fe86423688","shared_citers":3}],"time_series":[{"n":1,"year":2024},{"n":1,"year":2025},{"n":31,"year":2026}],"dependency_candidates":[]},"error":null,"updated_at":"2026-05-14T18:29:59.966710+00:00"},"identity_refresh":{"job_type":"identity_refresh","status":"succeeded","result":{"items":[{"title":"Qwen3 Technical Report","outcome":"unchanged","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","resolver":"local_arxiv","confidence":0.98,"old_work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e"}],"counts":{"fixed":0,"merged":0,"unchanged":1,"quarantined":0,"needs_external_resolution":0},"errors":[],"attempted":1},"error":null,"updated_at":"2026-05-14T18:29:47.516020+00:00"},"summary_claims":{"job_type":"summary_claims","status":"succeeded","result":{"title":"RouteLLM: Learning to Route LLMs with Preference Data","claims":[{"claim_text":"Large language models (LLMs) exhibit impressive capabilities across a wide range of tasks, yet the choice of which model to use often involves a trade-off between performance and cost. More powerful models, though effective, come with higher expenses, while less capable models are more cost-effective. To address this dilemma, we propose several efficient router models that dynamically select between a stronger and a weaker LLM during inference, aiming to optimize the balance between cost and response quality. We develop a training framework for these routers leveraging human preference data an","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks RouteLLM: Learning to Route LLMs with Preference Data because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-14T18:29:55.969575+00:00"}},"summary":{"title":"RouteLLM: Learning to Route LLMs with Preference Data","claims":[{"claim_text":"Large language models (LLMs) exhibit impressive capabilities across a wide range of tasks, yet the choice of which model to use often involves a trade-off between performance and cost. More powerful models, though effective, come with higher expenses, while less capable models are more cost-effective. To address this dilemma, we propose several efficient router models that dynamically select between a stronger and a weaker LLM during inference, aiming to optimize the balance between cost and response quality. We develop a training framework for these routers leveraging human preference data an","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks RouteLLM: Learning to Route LLMs with Preference Data because it crossed a citation-hub threshold.","role_counts":[]},"graph":{"co_cited":[{"title":"FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance","work_id":"5c567414-35de-4373-ae7a-1790da594e8d","shared_citers":18},{"title":"Huang, L., Yu, W., Ma, W., Zhong, W., Feng, Z., Wang, H., Chen, Q., Peng, W., Feng, X., Qin, B., and Liu, T","work_id":"ab0c1f58-f46a-4885-8afb-6a26e45608fb","shared_citers":8},{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":8},{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":7},{"title":"Hybrid llm: Cost-efficient and quality- aware query routing","work_id":"9dda325d-c570-4795-a95c-95c467dd09b4","shared_citers":7},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":6},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":6},{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":6},{"title":"AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation","work_id":"92b7eb9c-c3d8-4518-a376-06fa15dd895b","shared_citers":5},{"title":"Scaling Laws for Neural Language Models","work_id":"b7dd8749-9c45-4977-ab9b-64478dce1ae8","shared_citers":5},{"title":"Automix: Automatically mixing language models","work_id":"721159b7-a5b2-42e0-94e9-8aa5e1088d57","shared_citers":4},{"title":"MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework","work_id":"891b9780-a800-4e3c-bba0-53597ab8dc98","shared_citers":4},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":4},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":3},{"title":"Efficiently serving llm reasoning programs with certaindex","work_id":"57494d13-0d57-400a-8e38-67a61f1677d1","shared_citers":3},{"title":"Gemma: Open Models Based on Gemini Research and Technology","work_id":"a9ea2870-df28-40b8-a9e0-a7e9a116f793","shared_citers":3},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":3},{"title":"Graphrouter: A graph-based router for llm selections","work_id":"ed7f55c7-58fe-463b-af50-633c54486e8e","shared_citers":3},{"title":"Kimi k1.5: Scaling Reinforcement Learning with LLMs","work_id":"bff96ab1-bd6a-4585-be23-74fdb51969c7","shared_citers":3},{"title":"KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache","work_id":"735737c3-24e5-41c3-ab4f-04edcb36731c","shared_citers":3},{"title":"Pan, H., Tennenholtz, G., Mannor, S., Chi, C.-W., Brekel- mans, R., Shah, P., and Tewari, A","work_id":"86cd33f5-f0be-4d28-b775-10e356b31e74","shared_citers":3},{"title":"Qwen2.5 Technical Report","work_id":"d8432992-4980-4a81-85c7-9fa2c2b87f85","shared_citers":3},{"title":"ReAct: Synergizing Reasoning and Acting in Language Models","work_id":"407a2351-25f1-497d-b611-f77d0292a8e6","shared_citers":3},{"title":"Reward-guided speculative decoding for efficient llm reasoning","work_id":"460cb265-1bb5-4de5-a476-57fe86423688","shared_citers":3}],"time_series":[{"n":1,"year":2024},{"n":1,"year":2025},{"n":31,"year":2026}],"dependency_candidates":[]},"authors":[]}}