发表机构
Graduate School of Engineering, The University of Tokyo（东京大学大学院工学系研究科）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究提出测试时干预方法RoFB，将中间预测转为耦合力注入循环模型隐动态，在数独、迷宫任务的三类循环模型上提升性能，且计算成本相当或更低。
AI 中文摘要
循环模型通过共享计算模块反复更新隐状态，已成为解决复杂推理任务的强大架构。现有推理阶段方法通过增加运行步数或采样更多轨迹来扩展计算，但忽略了每条轨迹中揭示的信息。本文表明，可在推理阶段利用循环模型自身的读出概率引导隐动态，无需重新训练。我们提出读出反馈（Readout Feedback, RoFB），这是一种测试时干预方法，将中间预测转换为逐标记的成对耦合力，注入隐动态中。在数独和迷宫任务上的三种循环模型（AKOrN、ItrSA++、TRM）实验显示，RoFB在6个模型-任务对中的4个上取得了明显性能提升，达到仅增加步数或从多条轨迹中选择无法获得的性能，且计算成本相当或更低。这些结果表明，隐动态的闭环引导可作为循环推理模型的一种互补推理阶段控制机制。
英文摘要
Recurrent models, which repeatedly update latent states with shared computation blocks, have emerged as powerful architectures for solving complex reasoning tasks. Existing inference-time methods scale computation by running more steps or sampling more trajectories, but ignore information revealed within each trajectory. Here we show that recurrent models can be improved at inference time by using their own readout probabilities to steer latent dynamics without retraining. We introduce Readout Feedback (RoFB), a test-time intervention that converts intermediate predictions into token-wise pairwise coupling forces injected into the latent dynamics. Across three recurrent models (AKOrN, ItrSA++, TRM) on Sudoku and Maze, RoFB yields clear gains in four of six model-task pairs and a small gain in one. The four positive pairs achieve performance unattainable by merely running more steps or selecting from multiple trajectories, at comparable or lower computational cost. These results suggest that closed-loop steering of latent dynamics can serve as a complementary inference-time control mechanism for recurrent reasoning models.
Comments44 pages, 14 figures. Expanded analysis and experiments
