语言模型在隐空间自发形成搜索与回溯机制，提升推理准确性。
Emergent Search and Backtracking in Latent Reasoning Models
- 模型在隐状态中完成无文本的思维过程，逐步探索并修正答案。
- 32%的推理实例出现回溯，带来34%的准确率提升。
- 适合研究模型内部推理机制或改进生成式AI的可解释性。
当语言模型不依赖文字进行思考时会发生什么？标准推理模型通过思维链（chain-of-thought）输出中间步骤；而隐空间推理变换器（LRT）则完全在连续隐藏空间中进行推敲。我们在多选题问答基准上解码模型每一步的信念演化。发现模型自发学习到一种结构化的搜索过程：经历探索期（概率分布扩散）、对领先选项的初步认同、最终收敛或回溯。回溯现象普遍（32%实例），有益（比未回溯实例高出34%准确率），且主要从语义相近的干扰项转向正确答案。该搜索过程具有自适应性：将干扰项替换为不合理选项后，探索时间缩短54%。隐空间推理模型在激活空间实现了思维链通过语言达成的能力——犯错、察觉并纠正。
原文摘要 · Abstract (English)
What happens when a language model thinks without words? Standard reasoning LLMs verbalize intermediate steps as chain-of-thought; latent reasoning transformers (LRTs) instead perform deliberation entirely in continuous hidden space. We investigate an LRT, decoding the model's evolving beliefs at every step on a multiple-choice QA benchmark. We find that the model spontaneously learns a structured search process in latent space. Deliberation follows a consistent trajectory: an exploration phase where probability mass spreads across candidates, tentative commitment to a frontrunner, and either convergence or backtracking. Backtracking is prevalent (32% of instances), beneficial (34% accuracy gain over non-backtracking instances), and predominantly directed away from the semantically closest distractor toward the correct answer. The search is adaptive: replacing distractors with implausible alternatives shortens exploration by 54%. Latent reasoning models achieve in activation space what chain-of-thought achieves through words: the ability to be wrong, notice, and recover.
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。