aiqbench analysis
This paper is squarely about AI: it proposes streaming quantile matching, an optimal-transport-based post-hoc calibration method for aligning LLM-as-a-judge scores with human ratings. The main claims are that it preserves score ordinality and matches distributions better than point-wise calibration, and that calibration narrows the error gap between small and large judge models, which could make cheaper judges more viable. Readers should treat these as results from a specific evaluation framework, and check how they hold across tasks and anchor-set sizes before relying on them.
Quantiles, Not Points: Streaming Quantile Matching for Ordinal-Aware LLM-as-a-Judge Calibration
arXiv:2611.00447v1 Announce Type: new Abstract: LLM-as-a-judge scores are often distributionally shifted from human ratings on any given evaluation task, and post-hoc calibration (mapping LLM to human scores in a small labeled anchor set) has become a popular method to correct for this shift. In this paper we derive the calibration from two positions: that LLMaJ scores are ordinal ratings drawn from a distribution rather than isolated points. We present streaming quantile matching (SQM), a version of a classical method not previously bench marked for LLMaJ calibration, which is built on the two positions by computing the optimal-transport map between the LLM and human score distributions. We develop an evaluation framework and within it we show that unlike point-wise calibrations, SQM respects the ordinality of the human anchor set and achieves the best distributional fidelity and the top overall rank across models and data. We also show that in general, post-hoc calibration narrows the raw error gap between smaller and larger judge models, which opens the possibility of more efficient LLM judges.
Abstract:LLM-as-a-judge scores are often distributionally shifted from human ratings on any given evaluation task, and post-hoc calibration (mapping LLM to human scores in a small labeled anchor set) has become a popular method to correct for this shift. In this paper we derive the calibration from two positions: that LLMaJ scores are ordinal ratings drawn from a distribution rather than isolated points. We present streaming quantile matching (SQM), a version of a classical method not previously bench marked for LLMaJ calibration, which is built on the two positions by computing the optimal-transport map between the LLM and human score distributions. We develop an evaluation framework and within it we show that unlike point-wise calibrations, SQM respects the ordinality of the human anchor set and achieves the best distributional fidelity and the top overall rank across models and data. We also show that in general, post-hoc calibration narrows the raw error gap between smaller and larger judge models, which opens the possibility of more efficient LLM judges.
Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
arXiv:2611.00447 [cs.AI]
(or arXiv:2611.00447v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2611.00447 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 264 words