Recent advancements in large reasoning models (LRMs) have greatly improved their capabilities on complex reasoning tasks through Long Chains of Thought (CoTs). However, this approach often results in substantial redundancy, impairing computational efficiency and causing significant delays in real-time applications. Recent studies show that longer reasoning chains are frequently uncorrelated with correctness and can even be detrimental to accuracy. In a further in-depth analysis of this phenomenon, we surprisingly uncover and empirically verify that LRMs implicitly know the appropriate time to stop thinking, while this capability is obscured by current sampling paradigms. Motivated by this, we introduce SAGE (Self-Aware Guided Efficient Reasoning), a novel sampling paradigm that unleashes this efficient reasoning potential. Furthermore, integrating SAGE as mixed sampling into group-based reinforcement learning (SAGE-RL) enables SAGE-RL to effectively incorporate SAGE-discovered efficient reasoning patterns into standard pass@1 inference, markedly enhancing both the reasoning accuracy and efficiency of LRMs across multiple challenging mathematical benchmarks.
Your Models Have Thought Enough: Training Large Reasoning Models to Stop Overthinking
Shanghai Institute of Artificial Intelligence for Education, East China Normal University(华东师范大学上海智能教育研究院);School of Data Science, Fudan University(复旦大学数据科学学院);College of Computer Science and Artificial Intelligence, Fudan University(复旦大学计算机科学与人工智能学院);Antgroup(蚂蚁集团)
Single Headed Attention RNN: Stop Thinking With Your Head
Times Labs(时代实验室)
语言建模的主要方法都是痴迷于我年轻时的电视节目-即Transformers和芝麻街。Transformers this,Transformers that,and over here a bonfire worth of GPU-TPU-neuromorphic wafer scale silicon.我们选择了旧的和经过验证的技术的懒惰路径,并使用了一个受加密启发的首字母缩略词:单头注意力RNN(SHA-RNN)。作者唯一的目标是表明,如果我们不是痴迷于一个略有不同的缩写词和略有不同的结果,整个领域可能会朝着不同的方向发展。我们采用了一个以前只基于无聊的LSTM的强大语言模型,并在enwik 8上将其与最先进的字节级语言模型结果进行了一箭之遥的比较。这项工作没有经过密集的超参数优化,完全生活在一个商品台式机,使作者的小工作室公寓太温暖,在旧金山的夏天。最终的结果是可以实现的正负24小时内在一个单一的GPU作为作者是不耐烦的。注意力机制也很容易扩展到大的上下文,以最小的计算。就拿芝麻街来说
Exploring the System 1 Thinking Capability of Large Reasoning Models
Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所);School of Cyber Security, University of Chinese Academy of Sciences(中国科学院大学网络安全学院);Tongyi Lab, Alibaba Group(阿里巴巴集团通义实验室)
The Zero-Step Thinking: An Empirical Study of Mode Selection as Harder Early Exit in Reasoning Models
The Key Laboratory of Cognition and Decision Intelligence for Complex Systems, Institute of Automation, Chinese Academy of Sciences(认知与决策智能复杂系统重点实验室,自动化研究所,中国科学院);School of Artificial Intelligence, University of Chinese Academy of Sciences(人工智能学院,中国科学院大学);Shanghai Artificial Intelligence Laboratory(上海人工智能实验室)
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)
DART: Difficulty-Adaptive Reasoning Truncation for Efficient Large Language Models
School of Artificial Intelligence, Shanghai Jiao Tong University(上海交通大学人工智能学院);CSE Department, The Chinese University of Hong Kong(香港中文大学计算机科学与工程系);Research and Development Department, SIMMIR Tech(Simmir科技研发部);School of Information Science and Technology, Xiamen University Tan Kah Kee College(厦门大学信息科学与技术学院);The University of Hong Kong(香港大学)
Tree-of-Mixed-Thought: Combining Fast and Slow Thinking for Multi-hop Visual Reasoning
University of Science and Technology of China(中国科学技术大学);cmss.chinamobile.com(中移(苏州)软件技术有限公司);Lin Gang Laboratory(林冈实验室);ShanghaiTech University(上海科技大学)
FPT IS AI R&D Center(FPT IS AI研发中心);FPT Software AI Center(FPT软件AI中心);Kaopiz Solutions(Kaopiz解决方案公司);VNU - University of Engineering and Technology(越南国家大学工程技术大学)