发表机构
Shanghai Jiao Tong University; Taobao & Tmall Group（上海交通大学; 淘宝天猫集团）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对潜在推理缺乏过程监督导致表示坍缩的问题，提出原型中介过程监督（PMPS），通过原型空间软对齐和渐进序列对齐提供结构化监督，显著压缩输出长度并提升准确率。
AI 中文摘要
潜在推理方法通过用紧凑的连续空间嵌入替代冗长的显式思维链（CoT）标记，提升了标记级效率和鲁棒性。然而，现有方法缺乏对这些潜在嵌入的直接过程监督，这常常导致表示坍缩和信息分布不均。为解决此问题，我们提出了原型中介过程监督（PMPS），该方法引入可学习的推理原型作为语义锚点，为潜在推理提供结构化的过程级监督。PMPS将潜在嵌入和显式CoT嵌入投影到共享的原型空间中，通过原型分配实现不等长表示之间的多对多软对齐。同时，我们引入了一个渐进式序列对齐（PSA）模块以进一步指导训练：位置先验最初鼓励序列对齐结构，随后逐渐放松以允许自适应匹配。实验结果表明，在GSM8K-Aug上，PMPS将输出标记长度压缩至显式CoT的50%以下。与领先基线SIM-CoT相比，我们的方法在不同模型家族中平均准确率提升了2.08%。在GPT-2上，PMPS甚至超越了CoT-SFT。在更大模型和更具挑战性的任务上，PMPS在所有具有可比输出长度的潜在推理方法中持续取得最高准确率。
英文摘要
Latent reasoning approaches enhance token-level efficiency and robustness by replacing verbose, explicit chain-of-thought (CoT) tokens with compact continuous-space embeddings. However, existing methods lack efficient process supervision over these embeddings, which often leads to representation collapse and uneven information distribution. % However, supervising these latent embeddings is challenging because a fixed number of latent states must capture information from variable-length reasoning traces. To address this, we propose Prototype-Mediated Process Supervision (PMPS), which introduces learnable reasoning prototypes as semantic anchors to provide structural process-level supervision for latent reasoning. PMPS projects latent embeddings and explicit CoT embeddings into a shared prototype space, achieving many-to-many soft alignment between unequal-length representations through prototype assignment. Meanwhile, we introduce a Progressive Sequential Alignment (PSA) module to further guide training: positional priors initially encourage sequential alignment structure, then gradually relax to permit adaptive matching. Experimental results show that PMPS compresses output token length to under 50% of explicit CoT on GSM8K-Aug. Compared to leading baseline SIM-CoT, our method achieves average accuracy gains of 2.51% across different model families. On GPT-2, accuracy of PMPS even surpasses CoT-SFT. On larger models and a more challenging task, PMPS consistently attains the highest accuracy among all latent reasoning methods with comparable output length.
