# Decomposing LLM-Judge Uncertainty to Target Expert Labels

arXiv: https://arxiv.org/abs/2609.06444 · HTML: https://mlsanity.com/paper/2609.06444/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 11 of its 16 references that are in the map.
- First posted on arXiv 6 Sep 2026.

## Builds on (most cited first, 11 of 11)

- [The 'Problem' of Human Label Variation: On Ground Truth in Data, Modeling and Evaluation](https://mlsanity.com/paper/2211.02570/) (2022)
- [Trust or Escalate: LLM Judges with Provable Guarantees for Human Agreement](https://mlsanity.com/paper/2407.18370/) (2024)
- [Stop Measuring Calibration When Humans Disagree](https://mlsanity.com/paper/2210.16133/) (2022)
- [Limits to scalable evaluation at the frontier: LLM as Judge won't beat twice the data](https://mlsanity.com/paper/2410.13341/) (2024)
- [From Risk to Uncertainty: Generating Predictive Uncertainty Measures via Bayesian Estimation](https://mlsanity.com/paper/2402.10727/) (2024)
- [Analyzing Uncertainty of LLM-as-a-Judge: Interval Evaluations with Conformal Prediction](https://mlsanity.com/paper/2509.18658/) (2025)
- [Epistemic Uncertainty Sampling](https://mlsanity.com/paper/1909.00218/) (2019)
- [SCOPE: Selective Conformal Optimized Pairwise LLM Judging](https://mlsanity.com/paper/2602.13110/) (2026)
- [LeWiDi-2025 at NLPerspectives: Third Edition of the Learning with Disagreements Shared Task](https://mlsanity.com/paper/2510.08460/) (2025)
- [LLM Performance Predictors: Learning When to Escalate in Hybrid Human-AI Moderation Systems](https://mlsanity.com/paper/2601.07006/) (2026)
- [On Cost-Effective LLM-as-a-Judge Improvement Techniques](https://mlsanity.com/paper/2604.13717/) (2026)

## Abstract (arXiv)

An LLM judge evaluates outputs at scale. Experts should label only where it is least sure. Its natural escalation signal conflates two uncertainties: aleatoric, real disagreement in the expert pool, which labels cannot reduce, and epistemic, the judge's ignorance, which labels do reduce. A small Bayesian model separates them: a regression on labels already collected learns how far to trust a black-box judge's prediction. Both components follow as simple formulas, with no sampling or further judge calls. The components isolate on a real LLM judge against exactly known truth, and stated confidence is no guide to its actual error. On real human disagreement (ChaosNLI) the epistemic ranking removes 83% more error than total uncertainty for the same expert labels, though simply escalating the least-labelled items does as well there. We demonstrate we can estimate where a judge is ignorant rather than where experts genuinely disagree, and propose using this to direct expert labelling. Code and data are available at https://github.com/composo-ai/judge-uncertainty-decomposition.