不用参考答案也能准确评分大模型输出，靠的是模型内部的隐藏信号。
Reference-Free Rating of LLM Responses via Latent Information
- 从模型内部信号提取评分：概率加权、验证器概率、激活值线性探测
- 相比传统方法，评分更稳定，对齐度更高，能有效区分优劣响应
- 适合需要无参考评估的场景，如Best-of-N筛选和多教师蒸馏
在无参考的情况下，单次输出的大模型评分是否可靠？我们研究了让裁判模型对自由文本输出打分的常见做法，发现存在两大系统性问题：评分在采样时不稳定，且校准差，导致分数集中在量表顶端并频繁出现平局。为此，我们提出基于隐含信息的「潜伏裁判」（Latent Judges）方法，利用模型内部信号生成标量评分：(i) 基于整数评分的概率加权；(ii) 类似验证器的“是”概率；(iii) 在评分位置上训练的线性探测器。在多种成对与单评基准测试中，潜伏方法表现匹配或超越标准提示，尤其在成对准确率和列表排序方面有显著提升。概率加权得分在单评相关性上最强，而探测器在输出逻辑不校准时仍能恢复有效信号。结果表明，潜伏信息可提供确定且更具区分度的参考无关评估信号，有助于改进Best-of-N选择、多教师蒸馏与路由等策略。
原文摘要 · Abstract (English)
How reliable are single-response LLM-as-a-judge ratings without references, and can we obtain fine-grained, deterministic scores in this setting? We study the common practice of asking a judge model to assign Likert-scale scores to free-text responses and show two systematic issues: scores are unstable under sampling and poorly calibrated, leading to compression near the top of the scale and frequent ties. We then propose and evaluate Latent Judges, which derive scalar ratings from internal model signals: (i) probability-weighted scores over integer ratings, (ii) verifier-style probabilities of "yes", and (iii) linear probes trained on model activations at the rating position. Across a broad suite of pairwise and single-rating benchmarks, latent methods match or surpass standard prompting, with consistent gains on pairwise accuracy and listwise ranking relevant to Best-of-N selection. Probability-weighted scores achieve the strongest single-rating correlations, while probes recover useful signals when output logits are miscalibrated. These results indicate that latent information provides deterministic and more discriminative signals for reference-free evaluation, and can improve selection and training approaches like Best-of-$N$, multi-teacher distillation, and routing.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。