Latent reasoning models (LRMs) have attracted significant research interest due to their low inference cost (relative to explicit reasoning models) and theoretical ability to explore multiple reasoning paths in parallel. However, these benefits come at the cost of reduced interpretability: LRMs are difficult to monitor because they do not reason in natural language. This paper presents an investigation into LRM interpretability by examining two state-of-the-art LRMs. First, we find that latent reasoning tokens are often unnecessary for LRMs' predictions; on logical reasoning datasets, LRMs can almost always produce the same final answers without using latent reasoning at all. This underutilization of reasoning tokens may partially explain why LRMs do not consistently outperform explicit reasoning methods and raises doubts about the stated role of these tokens in prior work. Second, we demonstrate that when latent reasoning tokens are necessary for performance, we can decode gold reasoning traces up to 65-93% of the time for correctly predicted instances. This suggests LRMs often implement the expected solution rather than an uninterpretable reasoning process. Finally, we present a method to decode a verified natural language reasoning trace from latent tokens without knowing a gold reasoning trace a priori, demonstrating that it is possible to find a verified trace for a majority of correct predictions but only a minority of incorrect predictions. Our findings highlight that current LRMs largely encode interpretable processes, and interpretability itself can be a signal of prediction correctness.
Structured Contrastive Learning for Interpretable Latent Representations
Department of Electrical and Electronic Engineering, Imperial College London(帝国理工学院伦敦电子与电气工程系);Institute of Biomedical Engineering, Department of Engineering Science, University of Oxford(牛津大学生物医学工程研究所)
Latent Diffusion Energy-Based Model for Interpretable Text Modeling
UCLA(加州大学洛杉矶分校);Beijing Institute for General Artificial Intelligence(北京通用人工智能研究院);Salesforce Research(赛富时研究院);Google Brain(谷歌大脑);Peking University(北京大学);Tsinghua University(清华大学)
MIRACLE: Multi-task Learning based Interpretable Regulation of Autoimmune Diseases through Common Latent Epigenetics
University of Illinois Urbana Champaign(伊利诺伊大学厄巴纳-香槟分校);University of Michigan-Shanghai Jiao Tong University Joint Institute(上海交通大学密歇根学院);Shanghai Jiao Tong University(上海交通大学);University of Michigan(密歇根大学)
Differentiable Autoencoding Neural Operator for Interpretable and Integrable Latent Space Modeling
Department of Mechanical Engineering, University of Utah, Salt Lake City, UT, USA(犹他大学机械工程系);Imaging Institute, University of Utah, Salt Lake City, UT, USA(犹他大学成像研究所)
H-LDM: Hierarchical Latent Diffusion Models for Controllable and Interpretable PCG Synthesis from Clinical Metadata
School of Cyber Engineering Xidian University Xi'an, China(电子科技大学信息工程学院西安中国);School of Telecommunications Engineering Xidian University Xi'an, China(电子科技大学电信工程学院西安中国)
Multiresolution Tensor Learning for Efficient and Interpretable Spatial Analysis
Khoury College of Computer Sciences, Northeastern University(东北大学库里计算机科学学院);Salesforce AI Research(Salesforce 人工智能研究院);California Institute of Technology(加州理工学院);Department of Computing and Mathematical Sciences, California Institute of Technology(加州理工学院计算与数学科学系);Computer Science and Engineering, University of California, San Diego(加州大学圣迭戈分校计算机科学与工程系)
Graph-in-Graph (GiG): Learning interpretable latent graphs in non-Euclidean domain for biological and healthcare applications
Technical University of Munich(慕尼黑工业大学);Ca’ Foscari University of Venice(威尼斯大学);USI University of Lugano(瑞士意大利语区大学(卢加诺大学));NVIDIA(英伟达);Johns Hopkins University(约翰斯·霍普金斯大学);University of Oxford(牛津大学)
Towards Interpretable Natural Language Understanding with Explanations as Latent Variables
Beihang University(北京航空航天大学);Tsinghua University(清华大学);South China University of Technology(华南理工大学);Sun Yat-sen University(中山大学);Microsoft Research(微软研究院);Mila-Québec AI Institute(Mila-魁北克人工智能研究所);HEC Montréal(蒙特利尔高等商学院)
Beijing Institute for General Artificial Intelligence(北京通用人工智能研究院);Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);School of Artificial Intelligence, Beijing University of Posts and Telecommunications(北京邮电大学人工智能学院);School of Artificial Intelligence for Science, Peking University(北京大学科学人工智能学院)
A Structural FHMM for Interpretable Disease Trajectories in T2DM
Swiss Data Science Center (SDSC)(瑞士数据科学中心(SDSC));Ecole Polytechnique Fédérale de Lausanne (EPFL)(洛桑联邦理工学院);ETH Zürich(苏黎世联邦理工学院);Institute of Pharmaceutical Sciences(药学院)
SLIM: Sparse Latent Steering for Interpretable and Property-Directed LLM-Based Molecular Editing
The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州));Nanjing University of Aeronautics and Astronautics(南京航空航天大学);The 63rd Research Institute, National University of Defense Technology, Nanjing(国防科技大学第六三研究所,南京)
Scientifically-Interpretable Reasoning Network (ScIReN): Discovering Hidden Relationships in the Carbon Cycle and Beyond
Department of Informatics and Intelligent Systems, Institute of Energy and the Environment, The Pennsylvania State University(信息学与智能系统系,能源与环境研究所,宾夕法尼亚州立大学)
Fractal Flow: Hierarchical and Interpretable Normalizing Flow via Topic Modeling and Recursive Strategy
School of Mathematics Harbin Institute of Technology(哈尔滨工业大学数学学院);School of Mathematics and Institute for Artificial Intelligence Harbin Instiute of Technology(哈尔滨工业大学数学学院和人工智能研究院);School of Earth and Space Sciences and Institute for Artificial Intelligence Peking University(北京大学地球和空间科学学院和人工智能研究院)