| title | Self-Evaluation Is Already There: Eliciting Latent Judge Calibration in Base LLMs with Minimal Data |
|---|---|
| authors | XiuYu Zhang, Yi Shan, Junfeng Fang, Zhenkai Liang |
| date | 2026-06-03 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2606.05122 |
| affiliation | National University of Singapore |
| upvotes | 2 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Large language models are increasingly evaluated by other models, raising a natural question: can a model predict how a judge will score its own output? We find that the ability is largely present before any targeted training: prompted few-shot, a base model already predicts an external judge's multi-attribute quality scores on open-ended responses well above chance across three benchmarks. We introduce Self-Evaluation Elicitation (SEE), a method that surfaces this latent ability through a short cycle comprising a calibration-coupled reinforcement learning phase that improves the answer and predicts the judge, followed by a masked distillation phase that sharpens the prediction while leaving the answer untouched. From 160 unique examples, roughly 31x fewer than a reinforcement learning baseline, SEE improves held-out calibration across three benchmarks while preserving answer quality. The elicited self-evaluation is sharply localized within the model's own token distribution and stable across judges it was never trained against, indicating a transferable notion of quality rather than a single judge's preference. These results reframe judge-aligned self-evaluation as a problem of elicitation rather than acquisition. |
cite
@misc{zhang2026self,
title={Self-Evaluation Is Already There: Eliciting Latent Judge Calibration in Base LLMs with Minimal Data},
author={XiuYu Zhang and Yi Shan and Junfeng Fang and Zhenkai Liang},
year={2026},
eprint={2606.05122},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2606.05122}
}