层间的真相：利用层内局部信息评分进行大型语言模型的不确定性估计

Large language models (LLMs) are often confidently wrong, making reliable uncertainty estimation (UE) essential. Output-based heuristics are inexpensive but fragile, while probing internal representations is effective yet high-dimensional and difficult to transfer. A compact, per-instance UE method is proposed that scores cross-layer agreement patterns in internal representations using a single forward pass. Across three models, this method matches probing in-distribution, with mean diagonal differences of at most -1.8 AUPRC percentage points and +4.9 Brier score points. Under cross-dataset transfer, it consistently outperforms probing, achieving off-diagonal gains of up to +2.86 AUPRC and +21.02 Brier points. Under 4-bit weight-only quantization, it remains effective.