Large language models (LLMs) achieve strong performance when all task-relevant information is available upfront, as in static prediction and instruction-following problems. However, many real-world decision-making tasks are inherently online: crucial information must be acquired through interaction, feedback is delayed, and effective behavior requires balancing information collection and exploitation over time. While in-context learning enables adaptation without weight updates, existing LLMs often struggle to reliably leverage in-context interaction experience in such settings. In this work, we show that this limitation can be addressed through training. We introduce ORBIT, a multi-task, multi-episode meta-reinforcement learning framework that trains LLMs to learn from interaction in context. After meta-training, a relatively small open-source model (Qwen3-14B) demonstrates substantially improved in-context online learning on entirely unseen environments, matching the performance of GPT-5.2 and outperforming standard RL fine-tuning by a large margin. Scaling experiments further reveal consistent gains with model size, suggesting significant headroom for learn-at-inference-time decision-making agents. Code reproducing the results in the paper can be found at https://github.com/XiaofengLin7/ORBIT.
Technion - Israel Institute of Technology（以色列理工学院）
We study meta-learning for adversarial multi-armed bandits. We consider the online-within-online setup, in which a player (learner) encounters a sequence of multi-armed bandit episodes. The player's performance is measured as regret against the best arm in each episode, according to the losses generated by an adversary. The difficulty of the problem depends on the empirical distribution of the per-episode best arm chosen by the adversary. We present an algorithm that can leverage the non-uniformity in this empirical distribution, and derive problem-dependent regret bounds. This solution compri…
In this work, we analyze the reinstatement mechanism introduced by Ritter et al. (2018) to reveal two classes of neurons that emerge in the agent's working memory (an epLSTM cell) when trained using episodic meta-RL on an episodic variant of the Harlow visual fixation task. Specifically, Abstract neurons encode knowledge shared across tasks, while Episodic neurons carry information relevant for a specific episode's task.
There has been rapidly growing interest in meta-learning as a method for increasing the flexibility and sample efficiency of reinforcement learning. One problem in this area of research, however, has been a scarcity of adequate benchmark tasks. In general, the structure underlying past benchmarks has either been too simple to be inherently interesting, or too ill-defined to support principled analysis. In the present work, we introduce a new benchmark for meta-RL research, emphasizing transparency and potential for in-depth analysis as well as structural richness. Alchemy is a 3D video game, i…
Retrieval-Augmented Decision Transformer: External Memory for In-context RL
ELLIS Unit, LIT AI Lab, Institute for Machine Learning, JKU Linz, Austria（ELLIS单位、LIT人工智能实验室、机器学习研究所、JKU林茨、奥地利）；Extensity AI；Google DeepMind（谷歌DeepMind）；Mila - Québec AI Institute（魁北克人工智能研究所）；NXAI
University of Science and Technology of China（中国科学技术大学）
The performance of deep reinforcement learning methods prone to degenerate when applied to environments with non-stationary dynamics. In this paper, we utilize the latent context recurrent encoders motivated by recent Meta-RL materials, and propose the Latent Context-based Soft Actor Critic (LC-SAC) method to address aforementioned issues. By minimizing the contrastive prediction loss function, the learned context variables capture the information of the environment dynamics and the recent behavior of the agent. Then combined with the soft policy iteration paradigm, the LC-SAC method alternate…
A Sentiment Consolidation Framework for Meta-Review Generation
School of Computing and Information Systems, The University of Melbourne（墨尔本大学计算与信息系统学院）；Language Technologies Institute, Carnegie Mellon University（卡内基梅隆大学语言技术研究所）
We propose and address a novel few-shot RL problem, where a task is characterized by a subtask graph which describes a set of subtasks and their dependencies that are unknown to the agent. The agent needs to quickly adapt to the task over few episodes during adaptation phase to maximize the return in the test phase. Instead of directly learning a meta-policy, we develop a Meta-learner with Subtask Graph Inference(MSGI), which infers the latent parameter of the task by interacting with the environment and maximizes the return given the latent parameter. To facilitate learning, we adopt an intri…
Reward decomposition is a critical problem in centralized training with decentralized execution~(CTDE) paradigm for multi-agent reinforcement learning. To take full advantage of global information, which exploits the states from all agents and the related environment for decomposing Q values into individual credits, we propose a general meta-learning-based Mixing Network with Meta Policy Gradient~(MNMPG) framework to distill the global hierarchy for delicate reward decomposition. The excitation signal for learning global hierarchy is deduced from the episode reward difference between before an…
Sample efficiency has been a key issue in reinforcement learning (RL). An efficient agent must be able to leverage its prior experiences to quickly adapt to similar, but new tasks and situations. Meta-RL is one attempt at formalizing and addressing this issue. Inspired by recent progress in meta-RL, we introduce BIMRL, a novel multi-layer architecture along with a novel brain-inspired memory module that will help agents quickly adapt to new tasks within a few episodes. We also utilize this memory module to design a novel intrinsic reward that will guide the agent's exploration. Our architectur…
Despite ample motivation from costly exploration and limited trajectory data, rapidly adapting to new environments with few-shot reinforcement learning (RL) can remain a challenging task, especially with respect to personalized settings. Here, we consider the problem of recommending optimal policies to a set of multiple entities each with potentially different characteristics, such that individual entities may parameterize distinct environments with unique transition dynamics. Inspired by existing literature in meta-learning, we extend previous work by focusing on the notion that certain envir…
Recent research considers few-shot intent detection as a meta-learning problem: the model is learning to learn from a consecutive set of small tasks named episodes. In this work, we propose ProtAugment, a meta-learning algorithm for short texts classification (the intent detection task). ProtAugment is a novel extension of Prototypical Networks, that limits overfitting on the bias introduced by the few-shots classification objective at each episode. It relies on diverse paraphrasing: a conditional language model is first fine-tuned for paraphrasing, and diversity is later introduced at the dec…
University of Michigan（密歇根大学）；Ford Motor Company（福特汽车公司）
It is essential for an automated vehicle in the field to perform discretionary lane changes with appropriate roadmanship - driving safely and efficiently without annoying or endangering other road users - under a wide range of traffic cultures and driving conditions. While deep reinforcement learning methods have excelled in recent years and been applied to automated vehicle driving policy, there are concerns about their capability to quickly adapt to unseen traffic with new environment dynamics. We formulate this challenge as a multi-Markov Decision Processes (MDPs) adaptation problem and dev…
An Adaptable Budget Planner for Enhancing Budget-Constrained Auto-Bidding in Online Advertising
School of Computer Science, Peking University（北京大学计算机学院）；Center on Frontiers of Computing Studies, Peking University（北京大学计算科学前沿中心）；Alibaba Group（阿里巴巴集团）
National Tsing Hua University（台湾清华大学）；Google AI Research（谷歌人工智能研究院）；Appier Inc.（沛星互动科技股份有限公司）
Designing an effective communication mechanism among agents in reinforcement learning has been a challenging task, especially for real-world applications. The number of agents can grow or an environment sometimes needs to interact with a changing number of agents in real-world scenarios. To this end, a multi-agent framework needs to handle various scenarios of agents, in terms of both scales and dynamics, for being practical to real-world applications. We formulate the multi-agent environment with a different number of agents as a multi-tasking problem and propose a meta reinforcement learning…
School of Computer Science and Engineering, Northeastern University（东北大学计算机科学与工程学院）；Department of Computer Science and Technology, Tsinghua University（清华大学计算机科学与技术系）；Alibaba Group（阿里巴巴集团）
Architecture, Dataset and Model-Scale Agnostic Data-free Meta-Learning
Tsinghua Shenzhen International Graduate School（清华大学深圳国际研究生院）；JD Explore Academy（京东探索研究院）；State University of New York at Buffalo（纽约州立大学布法罗分校）；The University of Sydney（悉尼大学）
MetaExp: Learning to Self-Improve through Experience via Verifiable Cross-Task Environment Scaling
University of Macau（澳门大学）；Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences（中国科学院深圳先进技术研究院）；Georgia Institute of Technology（佐治亚理工学院）；Alibaba Token Hub, Alibaba Group（阿里巴巴集团阿里 Token 中心）
This paper briefly reviews the connections between meta-learning and self-supervised learning. Meta-learning can be applied to improve model generalization capability and to construct general AI algorithms. Self-supervised learning utilizes self-supervision from original data and extracts higher-level generalizable features through unsupervised pre-training or optimization of contrastive loss objectives. In self-supervised learning, data augmentation techniques are widely applied and data labels are not required since pseudo labels can be estimated from trained models on similar tasks. Meta-le…
