Quantitative LLM Judges
This work addresses the challenge of aligning LLM evaluations with human judgments in a computationally efficient way, though it is incremental as it builds on existing LLM-as-a-judge frameworks.
The paper tackles the problem of LLM judges struggling to predict human preferences and numeric scores by proposing quantitative LLM judges that align evaluation scores to humans using regression models, resulting in improved predictive power as validated on four datasets.
LLM-as-a-judge is a framework where a large language model (LLM) evaluates the output of another LLM. While LLMs excel at producing qualitative textual evaluations, they often struggle to predict human preferences and numeric scores. We propose quantitative LLM judges, which align evaluation scores of existing LLM judges to humans in a given domain using regression models. The models are trained to improve the score of the original judge using its rationale and score. We present four quantitative judges for different types of absolute and relative feedback, which showcases the generality and versatility of our framework. Our framework is more computationally efficient than supervised fine-tuning and can be more statistically efficient when human feedback is limited, which is expected in practice. We validate these claims empirically on four datasets using two base judges. Our experiments show that quantitative judges can improve the predictive power of existing judges through post-hoc modeling.