Reasoning models enhance problem-solving by scaling test-time compute, yet they face a critical paradox: excessive thinking tokens often degrade performance rather than improve it. We attribute this to a fundamental architectural flaw: standard LLMs operate as "malloc-only" engines, continuously accumulating valid and redundant steps alike without a mechanism to prune obsolete information. To break this cycle, we propose Free()LM, a model that introduces an intrinsic self-forgetting capability via the Free-Module, a plug-and-play LoRA adapter. By iteratively switching between reasoning and cleaning modes, Free()LM dynamically identifies and prunes useless context chunks, maintaining a compact and noise-free state.
Extensive experiments show that Free()LM provides consistent improvements across all model scales (8B to 685B). It achieves a 3.3% average improvement over top-tier reasoning baselines, even establishing a new SOTA on IMOanswerBench using DeepSeek V3.2-Speciale. Most notably, in long-horizon tasks where the standard Qwen3-235B-A22B model suffers a total collapse (0% accuracy), Free()LM restores performance to 50%. Our findings suggest that sustainable intelligence requires the freedom to forget as much as the power to think.
MLLM-CTBench: A Benchmark for Continual Instruction Tuning with Reasoning Process Diagnosis
Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);Southeast University(东南大学);School of Computing, National University of Singapore(新加坡国立大学计算机学院)
Towards Reasoning-Preserving Unlearning in Multimodal Large Language Models
Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)(Mohamed bin Zayed人工智能大学);University of Queensland(昆士兰大学);Provable Responsible AI and Data Analytics (PRADA) Lab(可证明责任人工智能与数据分析实验室);King Abdullah University of Science and Technology (KAUST)(卡塔尔科学与技术大学)
Mila - Quebec AI Institute(魁北克人工智能研究所);Polytechnique Montréal(蒙特利尔理工学院);Université de Montréal(蒙特利尔大学);McGill University(麦吉尔大学);CIFAR AI Chair(CIFAR人工智能主席);Google DeepMind(谷歌DeepMind)
Mitigating Visual Forgetting via Take-along Visual Conditioning for Multi-modal Long CoT Reasoning
School of Artificial Intelligence, Nanjing University(南京大学人工智能学院);National Key Laboratory for Novel Software Technology, Nanjing University(南京大学新型软件技术国家重点实验室);Tencent(腾讯);Center for Language AI Research, Tohoku University(东北大学语言人工智能研究中心);RIKEN Center for Advanced Intelligence Project(日本理化学研究所先进人工智能项目中心)
"The Whole Is Greater Than the Sum of Its Parts": A Compatibility-Aware Multi-Teacher CoT Distillation Framework
State Key Laboratory of Human-Machine Hybrid Augmented Intelligence, Institute of Artificial Intelligence and Robotics, Xi’an Jiaotong University(人机混合增强智能国家重点实验室,人工智能与机器人研究院,西安交通大学);Nankai University(南开大学);The Hong Kong University of Science and Technology(Guangzhou)(香港科技大学(广州));School of Software Engineering, State Key Laboratory of Human-Machine Hybrid Augmented Intelligence, Institute of Artificial Intelligence and Robotics, Xi’an Jiaotong University(软件学院,人机混合增强智能国家重点实验室,人工智能与机器人研究院,西安交通大学)
Does Math Reasoning Improve General LLM Capabilities? Understanding Transferability of LLM Reasoning
Carnegie Mellon University(卡内基梅隆大学);University of Pennsylvania(宾夕法尼亚大学);University of Washington(华盛顿大学);M-A-P;The Hong Kong Polytechnic University(香港理工大学)
TopoBench: Benchmarking LLMs on Hard Topological Reasoning
Intercom Research(Intercom研究院);University College Dublin(都柏林大学);University of Galway(Galway大学);Salk Institute(Salk研究所);Dublin City University(都柏林城市大学);Trinity College Dublin(都柏林三一学院)
EEPO: Exploration-Enhanced Policy Optimization via Sample-Then-Forget
The Chinese University of Hong Kong(香港中文大学);Microsoft Research Asia(微软亚洲研究院);Hong Kong Baptist University(香港浸会大学);LMU Munich(慕尼黑大学);MoE Key Laboratory of High Confidence Software Technologies(教育部高可信软件技术重点实验室)
RAVEN: Robust Advertisement Video Violation Temporal Grounding via Reinforcement Reasoning
Tencent(腾讯公司);Department of Computer Science and Technology, Tsinghua University(清华大学计算机科学与技术系);Zhejiang University(浙江大学);Singapore University of Technology and Design(新加坡科技设计大学)
Seek in the Dark: Reasoning via Test-Time Instance-Level Policy Gradient in Latent Space
Institute for Artificial Intelligence, Peking University(北京大学人工智能研究院);NLCo Lab, Beijing Institute for General Artificial Intelligence(北京通用人工智能研究院NLCo实验室);Department of Automation, Tsinghua University(清华大学自动化系);Shanghai Jiao Tong University(上海交通大学);Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);University of California, Los Angeles(加州大学洛杉矶分校)
MIND: From Passive Mimicry to Active Reasoning through Capability-Aware Multi-Perspective CoT Distillation
State Key Laboratory of Human-Machine Hybrid Augmented Intelligence, National Engineering Research Center for Visual Information and Applications, and Institute of Artificial Intelligence and Robotics, Xi’an Jiaotong University(西安交通大学人机混合增强智能国家重点实验室、视觉信息与应用国家工程研究中心、人工智能与机器人研究所);Nankai University(南开大学);The Hong Kong University of Science and Technology(Guangzhou)(香港科技大学(广州));School of Software Engineering, Xi’an Jiaotong University(西安交通大学软件学院)
Unveiling the Magic of Code Reasoning through Hypothesis Decomposition and Amendment
University of Science and Technology of China(中国科学技术大学);Institute of Artificial Intelligence, Hefei Comprehensive National Science Center(合肥综合性国家科学中心人工智能研究院)
IBISAgent: Reinforcing Pixel-Level Visual Reasoning in MLLMs for Universal Biomedical Object Referring and Segmentation
Zhejiang University(浙江大学);Shanghai Artificial Intelligence Laboratory(上海人工智能实验室);Ningbo Global Innovation Center, Zhejiang University(浙江大学宁波全球创新中心);Zhejiang Key Laboratory of Digital-Intelligence Service Technology(浙江省数字智能服务技术重点实验室)
R$^{2}$Seg: Training-Free OOD Medical Tumor Segmentation via Anatomical Reasoning and Statistical Rejection
Carnegie Mellon University(卡内基梅隆大学);University of Cambridge(剑桥大学);Zhejiang University(浙江大学);ETH Zurich(苏黎世联邦理工学院);University of Illinois Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校)