Publication | Open Access
Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
424
Citations
0
References
2023
Year
Artificial IntelligenceLlm Fine-tuningChatbotEngineeringCommunicationLarge Language ModelCorpus LinguisticsText MiningLarge Language ModelsNatural Language ProcessingInformation RetrievalComputational LinguisticsLanguage StudiesHuman ComputationMachine TranslationConversational Recommender SystemComputer ScienceJudgement AggregationLlm-based AgentStrong LlmsChatbot ArenaChat AssistantsHuman-computer InteractionLinguistics
Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. The study investigates using strong LLMs as judges to evaluate chat assistants on open‑ended questions. The authors examine LLM‑as‑a‑judge’s biases and reasoning limits, propose mitigation strategies, and validate agreement with human preferences using MT‑bench and Chatbot Arena. Strong LLM judges such as GPT‑4 achieve over 80% agreement with human preferences, demonstrating that LLM‑as‑a‑judge is a scalable, explainable proxy for costly human judgments, and the MT‑bench and Chatbot Arena benchmarks complement each other in evaluating LLaMA and Vicuna variants.
Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge.