Large Reasoning Models (LRMs) excel at complex reasoning tasks, but their efficiency is often hampered by overly verbose outputs. Prior steering methods attempt to address this issue by applying a single, global vector to hidden representations -- an approach grounded in the restrictive linear representation hypothesis. In this work, we introduce FlowSteer, a nonlinear steering method that goes beyond uniform linear shifts by learning a complete transformation between the distributions associated with verbose and concise reasoning. This transformation is learned via Flow Matching as a velocity field, enabling precise, input-dependent control over the model's reasoning process. By aligning steered representations with the distribution of concise-reasoning activations, FlowSteer yields more compact reasoning than the linear shifts. Across diverse reasoning benchmarks, FlowSteer demonstrates strong task performance and token efficiency compared to leading inference-time baselines. Our work demonstrates that modeling the full distributional transport with generative techniques offers a more effective and principled foundation for controlling LRMs.
Euphonium: Steering Video Flow Matching via Process Reward Gradient Guided Stochastic Dynamics
Huazhong University of Science and Technology(华中科技大学);Shanghai Jiao Tong University(上海交通大学);Tencent Hunyuan(腾讯混元);University of Chinese Academy of Sciences(中国科学院大学)
CoCoEmo: Composable and Controllable Human-Like Emotional TTS via Activation Steering
The University of Melbourne, Australia(墨尔本大学);Wuhan University, China(武汉大学);The Hong Kong University of Science(香港科学与技术大学);The University of Auckland, New Zealand(奥克兰大学);Monash University, Australia(莫纳什大学)
Discrete Guidance Matching: Exact Guidance for Discrete Flow Matching
School of Statistics, East China Normal University(华东师范大学统计学院);Department of Statistics, University of California, Los Angeles(加州大学洛杉矶分校统计系);Department of Industrial and Systems Engineering, University of Minnesota(明尼苏达大学工业与系统工程系);School of Data Science, The Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳)数据科学学院)
FLOWR.root: A flow matching based foundation model for joint multi-purpose structure-aware 3D ligand generation and affinity prediction
Machine Learning & Computational Sciences, Pfizer Worldwide R&D(辉瑞全球研发 机器学习与计算科学部);Computational Chemistry, Medicine Design, Pfizer Worldwide R&D(辉瑞全球研发 计算化学与医学设计部);Doctoral School of Medical and Health Sciences, Jagiellonian University Medical College(杰拉西利昂大学医学院医学与健康科学博士学院);Department of Physicochemical Drug Analysis, Faculty of Pharmacy, Jagiellonian University Medical College(杰拉西利昂大学医学院药物物理化学分析系);Institute of Structural Biology, Molecular Targets and Therapeutics Center, Helmholtz Munich(海德堡-慕尼黑分子靶点与治疗中心 结构生物学研究所);TUM School of Natural Sciences, Department of Bioscience, Bayerisches NMR Zentrum, Technical University of Munich(慕尼黑技术大学自然科学学院 生物科学系,拜耳核磁共振中心)
Metacognitive Reuse: Turning Recurring LLM Reasoning Into Concise Behaviors
Meta;Mila-Quebec AI Institute, University of Montreal(魁北克AI研究所,蒙特利尔大学);Princeton University(普林斯顿大学);Mila--Quebec AI Institute, Université de Montréal(魁北克AI研究所,蒙特利尔大学)