[2603.28135] CoT2-Meta: Budgeted Metacognitive Control for Test-Time Reasoning

Computer Science > Artificial Intelligence arXiv:2603.28135 (cs) [Submitted on 30 Mar 2026] Title:CoT2-Meta: Budgeted Metacognitive Control for Test-Time Reasoning Authors:Siyuan Ma, Bo Gao, Zikai Xiao, Hailong Wang, Xinlei Yu, Rui Qian, Jiayu Qian, Luqi Gong, Yang Liu View a PDF of the paper titled CoT2-Meta: Budgeted Metacognitive Control for Test-Time Reasoning, by Siyuan Ma and 8 other authors View PDF HTML (experimental) Abstract:Recent test-time reasoning methods improve performance by generating more candidate chains or searching over larger reasoning trees, but they typically lack explicit control over when to expand, what to prune, how to repair, and when to abstain. We introduce CoT2-Meta, a training-free metacognitive reasoning framework that combines object-level chain-of-thought generation with meta-level control over partial reasoning trajectories. The framework integrates four components: strategy-conditioned thought generation, tree-structured search, an online process oracle for step-level reasoning evaluation, and a meta-controller that allocates computation through expansion, pruning, repair, stopping, and fallback decisions. Under matched inference budgets, CoT2-Meta consistently outperforms strong single-path, sampling-based, and search-based baselines, including ReST-MCTS. On the default backbone, it achieves 92.8 EM on MATH, 90.4 accuracy on GPQA, 98.65 EM on GSM8K, 75.8 accuracy on BBEH, 85.6 accuracy on MMMU-Pro, and 48.8 accuracy on HLE, with gain...