通过分析语言模型各层激活路径，提升不确定性判断准确性。
Reading Calibrated Uncertainty from Language Model Trajectories
- 提取11个几何特征追踪每层MLP更新路径
- 在选择性弃权任务中比MSP最高提升21 AURC点
- 可定位错误产生位置，适合需可信推理的场景
最大软最大化概率（MSP）是评估语言模型结构化输出不确定性量化的一种默认方法，虽然计算成本低，但常存在校准偏差。现有方法通过探测模型内部激活状态，将原始隐藏状态输入不透明分类器，将其视为静态快照，忽略了表征形成过程中的逐层轨迹。然而，相似终点可能源于不同路径，证据积累、强化或逆转的过程可能揭示最终概率所掩盖的不确定性。本文提取了11个尺度不变的几何特征，追踪每层MLP更新的累积路径，并输入稀疏线性探测器。该探测器在选择性弃权任务中优于MSP，性能提升随基线校准偏差增大而增加，最高达21 AURC点。由于每个特征均有明确几何意义，探测器系数可追踪错误在深度方向如何形成——哪些层过早下结论，哪些层与当前状态矛盾，以及路径如何偏离终点。
原文摘要 · Abstract (English)
The maximum softmax probability (MSP) represents a default approach when evaluating uncertainty quantification for language model generation with structured output. Although cheap, it is often miscalibrated. Methods that probe the model's internal activations feed raw hidden states into opaque classifiers, reading activations as static snapshots and leaving implicit the layer-wise trajectory by which a representation is formed. Yet, similar endpoints can arise from very different paths, and how evidence accumulates, reinforces, or reverses across depth might reveal uncertainty that final probabilities obscure. We extract eleven scale-invariant geometric features, tracing the cumulative path of per-layer MLP updates, and feed them to a sparse linear probe. The probe outperforms MSP under selective abstention, with gains scaling with baseline miscalibration up to 21 AURC points. Because every feature has a closed-form geometric meaning, the probe's coefficients trace how and where along depth errors take shape -- which layers commit prematurely, which contradict the running state, where trajectories drift away from their endpoint.
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。