← 研踪 PaperTrailarXiv:2609.38962 · v2
Alleviating Hallucination in Reasoning Tasks with Training-Free Uncertainty-Guided Steering
首次提交 2026-09-30 12:28
版本更新 2026-11-10 01:57
雷达观测 2026-11-10 15:22
摘要
Recent work on hallucination detection in large language models has shown that, for a fixed pre-trained model and reasoning task, it is possible to estimate the model's confidence in the correctness of its outputs. Such uncertainty estimates have primarily been used to improve truthfulness by detecting or filtering confabulations. In this work, we ask whether these signals can instead be used more proactively to directly improve the accuracy of model-generated answers. We propose USteer, a simple, training-free steering mechanism that adjusts a model's layer-wise activations during inference using the gradient of a confidence measure with respect to the activations. This procedure nudges generation toward outputs with lower uncertainty at inference time, without modifying model parameters or requiring additional supervision. We show that this approach consistently reduces hallucination across a range of reasoning tasks, demonstrating that confidence signals can be leveraged not only for detection, but also for effective inference-time control of model behavior.
作者备注：Neurips 2026 main conference paper
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v2 · 2026-11-10 01:57 · 最新观测版本
Alleviating Hallucination in Reasoning Tasks with Training-Free Uncertainty-Guided Steering
Recent work on hallucination detection in large language models has shown that, for a fixed pre-trained model and reasoning task, it is possible to estimate the model's confidence in the correctness of its outputs. Such uncertainty estimates have primarily been used to improve truthfulness by detecting or filtering confabulations. In this work, we ask whether these signals can instead be used more proactively to directly improve the accuracy of model-generated answers. We propose USteer, a simple, training-free steering mechanism that adjusts a model's layer-wise activations during inference using the gradient of a confidence measure with respect to the activations. This procedure nudges generation toward outputs with lower uncertainty at inference time, without modifying model parameters or requiring additional supervision. We show that this approach consistently reduces hallucination across a range of reasoning tasks, demonstrating that confidence signals can be leveraged not only for detection, but also for effective inference-time control of model behavior.
查看此版本解读 →查看此版本原文 ↗