编码、思考、解码：通过递归潜在思维扩展测试时推理
Encode, Think, Decode: Scaling test-time reasoning with recursive latent thoughts
AI总结：
本文提出编码-思考-解码（ETD）方法，通过在中期训练阶段迭代推理相关层来增强LLM推理能力，在17个基准上显著提升准确率，如GSM8K提升28.4%，MATH提升36%，并探索了自适应深度策略。
摘要：
提升大型语言模型（LLM）推理能力的大多数努力，要么涉及扩大参数数量和训练数据规模，要么通过让模型生成复杂的思维链来扩展推理计算量。受可解释性研究的启发，该研究表明推理任务所需的关键计算集中在有限范围的层中，我们引入了编码-思考-解码（ETD）方法，该方法通过在中期训练阶段训练基础模型在推理相关的小部分层上迭代，从而增强其推理能力。ETD在保留原始架构、参数数量、超参数和训练数据组成的同时，放大了潜在推理。在推理时对选定的层进行迭代，ETD模型在17个推理基准上取得了显著提升，包括在GSM8K上相对准确率提升28.4%，在MATH上使用OLMo-2 1B基础模型提升36%。我们还探索了一种自适应深度策略，该策略根据每个输入令牌调整计算量。我们的结果表明，递归潜在推理为更强的LLM推理提供了一条简单有效的路径。
英文摘要：
Most efforts to improve the reasoning capabilities of large language models (LLMs) involve either scaling the number of parameters and the size of training data, or scaling inference computation by letting models generate complex chains of thought. Motivated by interpretability studies showing that the crucial computation required for reasoning tasks is concentrated in a limited range of layers, we introduce Encode-Think-Decode (ETD), a method that enhances the reasoning capabilities of a base model by training it to iterate over a small subset of reasoning-relevant layers during the mid-training stage. ETD amplifies latent reasoning while preserving the original architecture, parameter count, hyperparameters, and training data composition. When iterating on the selected layers at inference time, ETD models yield substantial gains on 17 reasoning benchmarks, including +28.4% relative accuracy improvement on GSM8K and +36% on MATH with the OLMo-2 1B Base model. We also explore an adaptive depth strategy that adjusts the computation per input token. Our results show that recursive latent reasoning offers a simple and effective path to stronger LLM reasoning.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
