CLUE：基于隐状态聚类的经验式非参数验证方法
CLUE: Non-parametric Verification from Experience via Hidden-State Clustering
AI总结：
针对大语言模型输出质量评估的现有缺陷，本文提出非参数验证器CLUE，利用隐状态聚类判断输出正确性，在多个基准上显著优于基线方法。
摘要：
评估大语言模型（LLM）输出的质量是一项关键挑战。现有方法要么依赖文本层面的信息（如奖励模型、多数投票），这类方法容易过拟合到表面线索；要么依赖基于token概率的校准置信度，这类方法在校准度较差的模型上会失效。但实际上，这两类信号都只是更丰富信息源——模型内部隐状态——的局部投影：靠近token嵌入的早期层保留了支撑文本判断的语义和词汇特征，而后期层则逐渐与输出logits对齐，嵌入了与置信度相关的信息。\n本文直接探索将隐状态作为验证的统一基础。我们证明，解的正确性会在隐激活轨迹中编码为几何可分的特征标识。为验证这一点，我们提出了Clue（基于聚类与经验的验证），这是一种极简的非参数验证器。CLUE没有可训练参数，仅通过隐状态增量总结每条推理轨迹，并通过与基于过往经验形成的“成功”和“失败”聚类的最近质心距离来分类正确性。该方法的简洁性凸显了底层信号的强度。\n实验结果表明，在候选重排序任务中，CLUE的性能始终优于LLM-as-a-judge基线，且与现代基于置信度的方法相当或更优，在AIME 24/25和GPQA数据集上同时提升了top-1准确率和多数投票准确率。其中一个亮点是，在使用1.5B模型的AIME 24数据集上，CLUE将准确率从56.7%（64样本多数投票）提升至70.0%（16样本top多数投票）。
英文摘要：
Assessing the quality of Large Language Model (LLM) outputs presents a critical challenge. Previous methods either rely on text-level information (e.g., reward models, majority voting), which can overfit to superficial cues, or on calibrated confidence from token probabilities, which would fail on less-calibrated models. Yet both of these signals are, in fact, partial projections of a richer source of information: the model's internal hidden states. Early layers, closer to token embeddings, preserve semantic and lexical features that underpin text-based judgments, while later layers increasingly align with output logits, embedding confidence-related information. This paper explores hidden states directly as a unified foundation for verification. We show that the correctness of a solution is encoded as a geometrically separable signature within the trajectory of hidden activations. To validate this, we present Clue (Clustering and Experience-based Verification), a deliberately minimalist, non-parametric verifier. With no trainable parameters, CLUE only summarizes each reasoning trace by an hidden state delta and classifies correctness via nearest-centroid distance to ``success'' and ``failure'' clusters formed from past experience. The simplicity of this method highlights the strength of the underlying signal. Empirically, CLUE consistently outperforms LLM-as-a-judge baselines and matches or exceeds modern confidence-based methods in reranking candidates, improving both top-1 and majority-vote accuracy across AIME 24/25 and GPQA. As a highlight, on AIME 24 with a 1.5B model, CLUE boosts accuracy from 56.7% (majority@64) to 70.0% (top-maj@16).
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
