Lianmin Zheng, Wei-Lin Chiang, Ying Sheng
Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, Ion Stoica
2023 · arXiv
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.
Metadata: arXiv. Abstracts retain their original authors’ rights.
Original HN comments, with labels from simple text rules. A label describes the whole comment and may miss context or irony. Read the discussion to make up your own mind. About the labels
4 selected excerpts · 4 total mentions
Sorry, I should have been more specific. I meant LLMs are more reliable and accurate at judging than at generating from scratch. They basically achieve over 80% agreement with human evaluators [1]. This level of agreement is similar to the consensus rate between two human evaluators, making LLM-as-a-judge a scalable and reliable proxy for human judgment. [1] https://arxiv.org/abs/2306.05685 (2023)
For better or worse this has become the defacto standard in LLM Evaluation research papers since the LLM as a Judge paper [0] came out. Its also heavily embedded into frameworks like LangChain and LlamaIndex to evaluate RAG pipelines. [0] https://arxiv.org/abs/2306.05685 [1] https://arxiv.org/abs/2411.15594
Joe here. It's difficult to evaluate natural language responses that come from LLM applications - there are not hard metrics to measure performance like there are in say supervised machine learning tasks. For RAG, you have the response to evaluate as well as the retrieved context chunks. We found that using gpt-4 as an evaluator to measure the quality of RAG responses and the relevance of the context chunks gave similar results to using human evaluators at Tonic to do the same task. Some research also agrees that using LLMs as an evaluator for natural language tasks gives similar results to using human evaluators https://arxiv.org/abs/2306.05685. As far as whether using gpt-4 is a safe approach, the best you could ask for is that gpt-4's evaluations match those of human evaluators, and that's what we've found as well as this research.
There is one: "The agreement between GPT-4 and humans reaches 85%, which is even higher than the agreement among humans (81%). This means GPT-4’s judgments closely align with the majority of humans. We also show that GPT-4’s judgments may help humans make better judgments. During our data collection, when a human’s choice deviated from GPT-4, we presented GPT-4’s judgments to humans and ask if they are reasonable. Despite different views, humans deemed GPT-4’s judgments reasonable in 75% of cases and are even willing to change their choices in 34% of cases."[1] [1] https://arxiv.org/abs/2306.05685
More from the same topics.
Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan +11
Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition.
Tom B. Brown, Benjamin Mann, Nick Ryder +28
Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task.
A. Pluchino. A. E. Biondo, A. Rapisarda
The largely dominant meritocratic paradigm of highly competitive Western cultures is rooted on the belief that success is due mainly, if not exclusively, to personal qualities such as talent, intelligence, skills, efforts or risk taking.
David Manheim, Scott Garrabrant
There are several distinct failure modes for overoptimization of systems on the basis of metrics.