New method calibrates LLM-as-a-judge scores to human ratings using streaming quantile matching
A new arXiv paper proposes streaming quantile matching for ordinal-aware calibration of LLM-as-a-judge scores, addressing distributional shifts between model-generated scores and human ratings.
Read the full story at arXiv cs.AI ↗
Timeline · 1 report
- 2026-11-03 05:00 · arXiv cs.AI
Quantiles, Not Points: Streaming Quantile Matching for Ordinal-Aware LLM-as-a-Judge Calibration