让推理模型像生成一样试多种思路，提升复杂问题求解能力
Generative Recursive Reasoning
- 用随机潜变量路径替代单一确定推理轨迹
- 在逻辑推理和多解约束任务上优于传统递归模型
- 支持有条件推理与无条件生成，适合复杂决策场景
未来的神经推理系统如何实现扩展计算？递归推理模型（RRMs）通过共享转移函数进行迭代潜变量精炼，是自回归序列扩展的有前景替代方案。但现有RRMs多为确定性模型，仅沿单一潜变量轨迹收敛到唯一预测。本文提出生成式递归推理模型（GRAM），将递归潜变量推理转化为概率化多轨迹计算。GRAM将推理建模为随机潜变量路径，支持多重假设、不同求解策略及推理时通过递归深度和并行轨迹采样实现规模扩展。该框架构建了一个潜变量生成模型，可通过 $p_θ(y \ mid x)$ 实现条件推理，也可在固定或无输入下实现无条件生成 $p_θ(x)$。基于近似变分推断训练后，GRAM在结构化推理与多解约束满足任务上优于确定性递归与循环基线模型，并展现出无条件生成能力。
原文摘要 · Abstract (English)
How should future neural reasoning systems implement extended computation? Recursive Reasoning Models (RRMs) offer a promising alternative to autoregressive sequence extension by performing iterative latent-state refinement with shared transition functions. Yet existing RRMs are largely deterministic, following a single latent trajectory and converging to a single prediction. We introduce Generative Recursive reAsoning Models (GRAM), a framework that turns recursive latent reasoning into probabilistic multi-trajectory computation. GRAM models reasoning as a stochastic latent trajectory, enabling multiple hypotheses, alternative solution strategies, and inference-time scaling through both recursive depth and parallel trajectory sampling. This yields a latent-variable generative model supporting conditional reasoning via $p_θ(y \mid x)$ and, with fixed or absent inputs, unconditional generation via $p_θ(x)$. Trained with amortized variational inference, GRAM improves over deterministic recurrent and recursive baselines on structured reasoning and multi-solution constraint satisfaction tasks, while demonstrating an unconditional generation capability. https://ahn-ml.github.io/gram-website
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。