Energy-Based Reasoners：用能量函数引导循环推理，Sudoku-Extreme 准确率达 99.9%
Energy-Based Reasoners: Your Recurrent Reasoner Should Be an Energy-Based Model
新论文 Energy-Based Reasoners，把循环推理改成能量函数上的梯度下降，Sudoku-Extreme 拿到 99.9% 准确率，推理算力还省 345 倍，做多步推理的可以看看。
一篇 arXiv 论文提出 Energy-Based Reasoners（EBRs），在潜空间推理状态上学习能量函数，把循环更新表述为能量上的梯度下降。能量分数用于评估潜状态，其梯度引导状态精炼并给出停止循环的信号。训练上引入 exploration-based training，以发现更多模式并塑形更广状态集合上的能量景观。在四个算法推理基准上，EBRs 匹配或超过此前循环推理模型，Sudoku-Extreme 达 99.9% 准确率，推理计算比次优模型少 345 倍。在所有多解基准上，EBRs 实现不同有效解覆盖的最优水平，单样本准确率接近 100%。
Energy-Based Reasoners: Your Recurrent Reasoner Should Be an Energy-Based Model
Looping is a powerful inductive bias for reasoning. By repeatedly iterating a recurrent block, looped models can scale their effective depth and enable multi-step reasoning in latent space. However, existing looped models are not inherently guided by an explicit measure of progress, making it difficult to determine when to stop looping and how to improve upon the current state. To enable this, we introduce Energy-Based Reasoners (EBRs), which learn an energy function over latent reasoning states and formulate looped updates as gradient descent on this energy. The energy scores latent states, while its gradient guides refinement and provides a signal for stopping. To train EBRs effectively, we further introduce exploration-based training to discover additional modes and to better shape the energy landscape over a broader set of states. Across four algorithmic reasoning benchmarks, EBRs match or outperform prior recurrent reasoning models, achieving 99.9% accuracy on Sudoku-Extreme while using 345$\times$ less inference compute than the next-best model and maintaining comparable training compute. On all multi-solution benchmarks, EBRs achieve state-of-the-art coverage of distinct valid solutions while maintaining near-100% single-sample accuracy. These results suggest that energy-based models are a promising approach for strong reasoning performance.