Reinforcement learning (RL) agents often struggle to generalize to new tasks and contexts without updating their parameters, mainly because their learned representations and policies are overfit to the specifics of their training environments. To boost agents' in-context RL (ICRL) ability, this work formulates ICRL as a two-agent emergent communication problem and introduces CORAL (Communicative Representation for Adaptive RL), a framework that learns a transferable communicative context by decoupling latent representation learning from control. In CORAL, an Information Agent (IA) is pre-trained as a world model on a diverse distribution of tasks. Its objective is not to maximize task reward, but to build a world model and distill its understanding into concise messages. The emergent communication protocol is shaped by a novel Causal Influence Loss, which measures the effect that the message has on the next action. During deployment, the previously trained IA serves as a fixed contextualizer for a new Control Agent (CA), which learns to solve tasks by interpreting the provided communicative context. Our experiments demonstrate that this approach enables the CA to achieve significant gains in sample efficiency and successfully perform zero-shot adaptation with the help of pre-trained IA in entirely unseen sparse-reward environments, validating the efficacy of learning a transferable communicative representation.

There, demand is highly intermittent and bursty: long runs of zeros, with islands of high counts.

Let $S$ be a finite set, and $X_1,\ldots,X_n$ an i.i.d. uniform sample from $S$. To estimate the size $|S|$, without further structure, one can wait for repeats and use the birthday problem. This requires a sample size of the order $|S|^\frac{1}{2}$. On the other hand, if $S=\{1,2,\ldots,|S|\}$, the maximum of the sample blown up by $n/(n-1)$ gives an efficient estimator based on any growing sample size. This paper gives refinements that interpolate between these extremes. A general non-asymptotic theory is developed. This includes estimating the volume of a compact convex set, the unseen species problem, and a host of testing problems that follow from the question `Is this new observation a typical pick from a large prespecified population?' We also treat regression style predictors. A general theorem gives non-parametric finite $n$ error bounds in all cases.

We consider matrices $\boldsymbol{A}(\boldsymbolθ)\in\mathbb{R}^{m\times m}$ that depend, possibly nonlinearly, on a parameter $\boldsymbolθ$ from a compact parameter space $Θ$. We present a Monte Carlo estimator for minimizing $\text{trace}(\boldsymbol{A}(\boldsymbolθ))$ over all $\boldsymbolθ\inΘ$, and determine the sampling amount so that the backward error of the estimator is bounded with high probability. We derive two types of bounds, based on epsilon nets and on generic chaining. Both types predict a small sampling amount for matrices $\boldsymbol{A}(\boldsymbolθ)$ with small offdiagonal mass, and parameter spaces $Θ$ of small ``size.'' Dependence on the matrix dimension~$m$ is only weak or not explicit. The bounds based on epsilon nets are easier to evaluate and come with fully specified constants. In contrast, the bounds based on chaining depend on the Talagrand functionals which are difficult to evaluate, except in very special cases. Comparisons between the two types of bounds are difficult, although the literature suggests that chaining bounds can be superior.

We prove that the true underlying directed acyclic graph (DAG) in Gaussian linear structural equation models is identifiable as the minimum-trace DAG when the error variances are weakly increasing with respect to the true causal ordering. This result bridges two existing frameworks as it extends the identifiable cases within the minimum-trace DAG method and provides a principled interpretation of the algorithmic ordering search approach, revealing that its objective is actually to minimize the total residual sum of squares. On the computational side, we prove that the hill climbing algorithm with a random-to-random (R2R) neighborhood does not admit any strict local optima. Under standard settings, we confirm the result through extensive simulations, observing only a few weak local optima. Interestingly, algorithms using other neighborhoods of equal size exhibit suboptimal behavior, having strict local optima and a substantial number of weak local optima.

We introduce Unsupervised Partner Design (UPD) - a population-free, multi-agent reinforcement learning framework for robust ad-hoc teamwork that adaptively generates training partners without requiring pretrained partners or manual parameter tuning. UPD constructs diverse partners by stochastically mixing an ego agent's policy with biased random behaviours and scores them using a variance-based learnability metric that prioritises partners near the ego agent's current learning frontier. We show that UPD can be integrated with unsupervised environment design, resulting in the first method enabling fully unsupervised curricula over both level and partner distributions in a cooperative setting. Through extensive evaluations on Overcooked-AI and the Overcooked Generalisation Challenge, we demonstrate that this dynamic partner curriculum is highly effective: UPD consistently outperforms both population-based and population-free baselines as well as ablations. In a user study, we further show that UPD achieves higher returns than all baselines and was perceived as significantly more adaptive, more human-like, a better collaborator, and less frustrating.

Generative models, especially diffusion and flow-based models, have been promising in offline multi-agent reinforcement learning. However, integrating powerful generative models into this framework poses unique challenges. In particular, diffusion and flow-based policies suffer from low sampling efficiency due to their iterative generation processes, making them impractical in time-sensitive or resource-constrained settings. To tackle these difficulties, we propose OM2P (Offline Multi-Agent Mean-Flow Policy), a novel offline MARL algorithm to achieve efficient one-step action sampling. To address the misalignment between generative objectives and reward maximization, we introduce a reward-aware optimization scheme that integrates a carefully-designed mean-flow matching loss with Q-function supervision. Additionally, we design a generalized timestep distribution and a derivative-free estimation strategy to reduce memory overhead and improve training stability. Empirical evaluations on Multi-Agent Particle and MuJoCo benchmarks demonstrate that OM2P achieves superior performance, with up to a 3.8x reduction in GPU memory usage and up to a 10.8x speed-up in training time. Our approach represents the first to successfully integrate mean-flow model into offline MARL, paving the way for practical and scalable generative policies in cooperative multi-agent settings.

Autonomous driving systems face significant challenges in perceiving complex environments and making real-time decisions. Traditional modular approaches, while offering interpretability, suffer from error propagation and coordination issues, whereas end-to-end learning systems can simplify the design but face computational bottlenecks. This paper presents a novel approach to autonomous driving using deep reinforcement learning (DRL) that integrates bird's-eye view (BEV) perception for enhanced real-time decision-making. We introduce the \texttt{Mamba-BEV} model, an efficient spatio-temporal feature extraction network that combines BEV-based perception with the Mamba framework for temporal feature modeling. This integration allows the system to encode vehicle surroundings and road features in a unified coordinate system and accurately model long-range dependencies. Building on this, we propose the \texttt{ME$^3$-BEV} framework, which utilizes the \texttt{Mamba-BEV} model as a feature input for end-to-end DRL, achieving superior performance in dynamic urban driving scenarios. We further enhance the interpretability of the model by visualizing high-dimensional features through semantic segmentation, providing insight into the learned representations. Extensive experiments on the CARLA simulator demonstrate that \texttt{ME$^3$-BEV} outperforms existing models across multiple metrics, including collision rate and trajectory accuracy, offering a promising solution for real-time autonomous driving.

This work leverages adaptive social learning to estimate partially observable global states in multi-agent reinforcement learning (MARL) problems. Unlike existing methods, the proposed approach enables the concurrent operation of social learning and reinforcement learning. Specifically, it alternates between a single step of social learning and a single step of MARL, eliminating the need for the time- and computation-intensive two-timescale learning frameworks. Theoretical guarantees are provided to support the effectiveness of the proposed method. Simulation results verify that the performance of the proposed methodology can approach that of reinforcement learning when the true state is known.

Recently, multi-modal physiological signals based emotion recognition has garnered increasing attention in the field of brain-computer interfaces. Nevertheness, the associated multi-modal physiological features are often high-dimensional and inevitably include irrelevant, redundant, and noisy representation, which can easily lead to overfitting, poor performance, and high computational complexity in emotion classifiers. Feature selection has been widely applied to address these challenges. However, previous studies generally assumed that multi-modal physiological data are complete, whereas in reality, the data are often incomplete due to the openness of the acquisition and operational environment. For example, a part of samples are available in several modalities but not in others. To address this issue, we propose a novel method for incomplete multi-modal physiological signal feature selection called adaptive shared latent structure learning (ASLSL). Based on the property that similar features share similar emotional labels, ASLSL employs adaptive shared latent structure learning to explore a common latent space shared for incomplete multi-modal physiological signals and multi-dimensional emotional labels, thereby mitigating the impact of missing information and mining consensus information. Two most popular multi-modal physiological emotion datasets (DEAP and DREAMER) with multi-dimensional emotional labels were utilized to compare the performance between compare ASLSL and seventeen feature selection methods. Comprehensive experimental results on these datasets demonstrate the effectiveness of ASLSL.