Self-Evaluation Is Already There: Eliciting Latent Judge Calibration in Base LLMs with Minimal Data
2026-06-03 • Computation and Language
Computation and Language
AI summaryⓘ
The authors explored whether language models can guess how other models score their own responses. They found that even without special training, models can predict evaluation scores better than random on multiple tests. They created a method called Self-Evaluation Elicitation (SEE), which uses a two-step process to make these predictions more accurate and reliable, while keeping the original answers good. Their approach uses fewer examples than usual and works well on judges the models never saw before, suggesting the model learns a general idea of quality rather than copying one specific judge.
large language modelsmodel evaluationfew-shot promptingreinforcement learningdistillationmodel calibrationself-evaluationtoken distributionmulti-attribute scoringtransfer learning
Authors
XiuYu Zhang, Yi Shan, Junfeng Fang, Zhenkai Liang
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.