让大模型边输入边思考，提速推理且更准
Latent Reasoning with Supervised Thinking States
- 输入过程中生成思考标记，实时融合到后续输入中
- 数学题表现逼近链式思维，两跳问答延迟更低
- 适合需要高效推理的场景，如长序列状态追踪
链式思维（CoT）能提升大语言模型解决复杂任务的能力，但推理开销大。本文提出思维状态（Thinking States），在输入处理过程中生成思考标记序列，将这些思考转换为嵌入向量并注入后续输入。该方法利用思考标记的重复性特征，结合自然语言监督和教师强迫训练，实现并行化学习。实验表明，思维状态在多个推理任务上优于其他隐式推理方法，在数学题上接近链式思维表现，两跳问答任务中性能相当且延迟更低。在状态追踪任务中，其推理能力更强，能外推至训练时未见的更长序列。
原文摘要 · Abstract (English)
Reasoning with a chain-of-thought (CoT) enables Large Language Models (LLMs) to solve complex tasks but incurs significant inference costs due to the generation of long rationales. We propose Thinking States, a method that performs reasoning {\em while} the input is processing. Specifically, Thinking States generates sequences of thinking tokens every few input tokens, transforms the thoughts back into embedding space, and adds them to the following input tokens. This has two key advantages. First, it captures the recurrent nature of CoT, but where the thought tokens are generated as input is processing. Second, since the thoughts are represented as tokens, they can be learned from natural language supervision, and using teacher-forcing, which is parallelizable. Empirically, Thinking States outperforms other latent reasoning methods on multiple reasoning tasks, narrowing the gap to CoT on math problems, and matching its performance on 2-Hop QA with improved latency. On state-tracking tasks, we show Thinking States leads to stronger reasoning behavior than CoT, successfully extrapolating to longer sequences than seen during training.
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。