Soft Thinking：释放大语言模型在连续概念空间中的推理潜力
Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Space
- University of California, Santa Barbara（加州大学圣芭芭拉分校）
- University of California, Santa Cruz（加州大学圣克ruz分校）
- University of California, Los Angeles（加州大学洛杉矶分校）
- Purdue University（普渡大学）
- LMSYS Org（LMSYS组织）
- Microsoft（微软）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
针对当前大语言模型离散推理的局限性，研究提出无需训练的Soft Thinking方法，通过连续概念空间的软性概念词元实现类人推理，在数学与编程基准上提升了准确率且减少了词元用量，兼具可解释性。
AI中文摘要：
人类认知通常涉及通过抽象、流动的概念进行思考，而非严格使用离散的语言符号。然而，当前的推理模型受限于在人类语言的边界内进行推理，处理的是代表语义空间中固定点的离散词元嵌入。这种离散约束限制了此类推理模型的表达能力和上限潜力，往往导致推理路径探索不完整，因为标准的思维链（Chain-of-Thought, CoT）方法依赖于每一步采样一个词元。在本研究中，我们提出了Soft Thinking，这是一种无需训练的方法，通过在连续概念空间中生成软性的抽象概念词元，来模拟类人的“软性”推理。这些概念词元由词元嵌入的概率加权混合生成，由此构成连续概念空间，能够实现平滑过渡和更丰富的表征，超越传统的离散边界。从本质上讲，每个生成的概念词元都封装了来自相关离散词元的多重含义，能隐式探索多种推理路径，从而有效地收敛到正确答案。在多种数学和编程基准上的实证评估一致证明了Soft Thinking的有效性和高效性：与标准CoT相比，其pass@1准确率最高提升2.48个点，同时词元使用量最多减少22.4%。定性分析进一步表明，Soft Thinking的输出仍保持高度可解释性和可读性，凸显了其打破基于离散语言的推理由来已久的瓶颈的潜力。代码可在https://github.com/eric-ai-lab/Soft-Thinking获取。
英文摘要：
Human cognition typically involves thinking through abstract, fluid concepts rather than strictly using discrete linguistic tokens. Current reasoning models, however, are constrained to reasoning within the boundaries of human language, processing discrete token embeddings that represent fixed points in the semantic space. This discrete constraint restricts the expressive power and upper potential of such reasoning models, often causing incomplete exploration of reasoning paths, as standard Chain-of-Thought (CoT) methods rely on sampling one token per step. In this work, we introduce Soft Thinking, a training-free method that emulates human-like "soft" reasoning by generating soft, abstract concept tokens in a continuous concept space. These concept tokens are created by the probability-weighted mixture of token embeddings, which form the continuous concept space, enabling smooth transitions and richer representations that transcend traditional discrete boundaries. In essence, each generated concept token encapsulates multiple meanings from related discrete tokens, implicitly exploring various reasoning paths to converge effectively toward the correct answer. Empirical evaluations on diverse mathematical and coding benchmarks consistently demonstrate the effectiveness and efficiency of Soft Thinking, improving pass@1 accuracy by up to 2.48 points while simultaneously reducing token usage by up to 22.4% compared to standard CoT. Qualitative analysis further reveals that Soft Thinking outputs remain highly interpretable and readable, highlighting the potential of Soft Thinking to break the inherent bottleneck of discrete language-based reasoning. Code is available at https://github.com/eric-ai-lab/Soft-Thinking.
