However, controlling conventional multi-trap OT to achieve cooperative manipulation of multiple complex-shaped microrobots in dynamic environments poses a significant challenge. T o address this, we introduce Interactive OT Gym, a reinforcement learning (RL)-based simulation platform designed for OT -driven microrobotics. Our platform supports complex physical field simulations and integrates haptic feedback interfaces, RL modules, and context-aware shared control strategies tailored for OT -driven microrobot in cooperative biological object manipulation tasks. This integration allows for an adaptive blend of manual and autonomous control, enabling seamless transitions between human input and autonomous operation. We evaluated the effectiveness of our platform using a cell manipulation task. Experimental results show that our shared control system significantly improves micromanipulation performance, reducing task completion time by approximately 67% compared to using pure human or RL control alone and achieving a 100% success rate. With its high fidelity, interactivity, low cost, and high-speed simulation capabilities, Interactive OT Gym serves as a user-friendly training and testing environment for the development of advanced interactive OT -driven micromanipulation systems and control algorithms. For more details on the project, please see our website https://sites.google.com/view/otgym

We explore a method for improving the performance of large language models through self-reflection and reinforcement learning. By incentivizing the model to generate better self-reflections when it answers incorrectly, we demonstrate that a model's ability to solve complex, verifiable tasks can be enhanced even when generating synthetic data is infeasible and only binary feedback is available. Our framework operates in two stages: first, upon failing a given task, the model generates a self-reflective commentary analyzing its previous attempt; second, the model is given another attempt at the task with the self-reflection in context. If the subsequent attempt succeeds, the tokens generated during the self-reflection phase are rewarded. Our experimental results show substantial performance gains across a variety of model architectures, as high as 34.7% improvement at math equation writing and 18.1% improvement at function calling. Notably, smaller fine-tuned models (1.5 billion to 7 billion parameters) outperform models in the same family that are 10 times larger. Our novel paradigm is thus an exciting pathway to more useful and reliable language models that can self-improve on challenging tasks with limited external feedback.

Reinforcement learning (RL) in autonomous driving employs a trial-and-error mechanism, enhancing robustness in unpredictable environments. However, crafting effective reward functions remains challenging, as conventional approaches rely heavily on manual design and demonstrate limited efficacy in complex scenarios. To address this issue, this study introduces a responsibility-oriented reward function that explicitly incorporates traffic regulations into the RL framework. Specifically, we introduced a Traffic Regulation Knowledge Graph and leveraged Vision-Language Models alongside Retrieval-Augmented Generation techniques to automate reward assignment. This integration guides agents to adhere strictly to traffic laws, thus minimizing rule violations and optimizing decision-making performance in diverse driving conditions. Experimental validations demonstrate that the proposed methodology significantly improves the accuracy of assigning accident responsibilities and effectively reduces the agent's liability in traffic incidents.

Gene selection in high-dimensional genomic data is essential for understanding disease mechanisms and improving therapeutic outcomes. Traditional feature selection methods effectively identify predictive genes but often ignore complex biological pathways and regulatory networks, leading to unstable and biologically irrelevant signatures. Prior approaches, such as Lasso-based methods and statistical filtering, either focus solely on individual gene-outcome associations or fail to capture pathway-level interactions, presenting a key challenge: how to integrate biological pathway knowledge while maintaining statistical rigor in gene selection? To address this gap, we propose a novel two-stage framework that integrates statistical selection with biological pathway knowledge using multi-agent reinforcement learning (MARL). First, we introduce a pathway-guided pre-filtering strategy that leverages multiple statistical methods alongside KEGG pathway information for initial dimensionality reduction. Next, for refined selection, we model genes as collaborative agents in a MARL framework, where each agent optimizes both predictive power and biological relevance. Our framework incorporates pathway knowledge through Graph Neural Network-based state representations, a reward mechanism combining prediction performance with gene centrality and pathway coverage, and collaborative learning strategies using shared memory and a centralized critic component. Extensive experiments on multiple gene expression datasets demonstrate that our approach significantly improves both prediction accuracy and biological interpretability compared to traditional methods.

This paper studies the networked multi-agent reinforcement learning (NMARL) problem, where the objective of agents is to collaboratively maximize the discounted average cumulative rewards. Different from the existing methods that suffer from poor expression due to linear function approximation, we propose a distributed neural policy gradient algorithm that features two innovatively designed neural networks, specifically for the approximate Q-functions and policy functions of agents. This distributed neural policy gradient algorithm consists of two key components: the distributed critic step and the decentralized actor step. In the distributed critic step, agents receive the approximate Q-function parameters from their neighboring agents via a time-varying communication networks to collaboratively evaluate the joint policy. In contrast, in the decentralized actor step, each agent updates its local policy parameter solely based on its own approximate Q-function. In the convergence analysis, we first establish the global convergence of agents for the joint policy evaluation in the distributed critic step. Subsequently, we rigorously demonstrate the global convergence of the overall distributed neural policy gradient algorithm with respect to the objective function. Finally, the effectiveness of the proposed algorithm is demonstrated by comparing it with a centralized algorithm through simulation in the robot path planning environment.

Deep reinforcement learning (RL) agents frequently suffer from neuronal activity loss, which impairs their ability to adapt to new data and learn continually. A common method to quantify and address this issue is the tau-dormant neuron ratio, which uses activation statistics to measure the expressive ability of neurons. While effective for simple MLP-based agents, this approach loses statistical power in more complex architectures. To address this, we argue that in advanced RL agents, maintaining a neuron's learning capacity, its ability to adapt via gradient updates, is more critical than preserving its expressive ability. Based on this insight, we shift the statistical objective from activations to gradients, and introduce GraMa (Gradient Magnitude Neural Activity Metric), a lightweight, architecture-agnostic metric for quantifying neuron-level learning capacity. We show that GraMa effectively reveals persistent neuron inactivity across diverse architectures, including residual networks, diffusion models, and agents with varied activation functions. Moreover, resetting neurons guided by GraMa (ReGraMa) consistently improves learning performance across multiple deep RL algorithms and benchmarks, such as MuJoCo and the DeepMind Control Suite.

Vision-language models (VLMs) have achieved strong results on coding and math benchmarks that are challenging for humans, yet their ability to perform tasks that come naturally to humans--such as perception, spatial navigation, and memory management--remains understudied. Real video games are crafted to be intuitive for humans to learn and master by leveraging innate inductive biases, making them an ideal testbed for evaluating such capabilities in VLMs. To this end, we introduce VideoGameBench, a benchmark consisting of 10 popular video games from the 1990s that VLMs directly interact with in real-time. VideoGameBench challenges models to complete entire games with access to only raw visual inputs and a high-level description of objectives and controls, a significant departure from existing setups that rely on game-specific scaffolding and auxiliary information. We keep three of the games secret to encourage solutions that generalize to unseen environments. Our experiments show that frontier vision-language models struggle to progress beyond the beginning of each game. We find inference latency to be a major limitation of frontier models in the real-time setting; therefore, we introduce VideoGameBench Lite, a setting where the game pauses while waiting for the LM's next action. The best performing model, Gemini 2.5 Pro, completes only 0.48% of VideoGameBench and 1.6% of VideoGameBench Lite. We hope that the formalization of the human skills mentioned above into this benchmark motivates progress in these research directions.

Modelling bounded rational decision-making through information constrained processing provides a principled approach for representing departures from rationality within a reinforcement learning framework, while still treating decision-making as an optimization process. However, existing approaches are generally based on Entropy, Kullback-Leibler divergence, or Mutual Information. In this work, we highlight issues with these approaches when dealing with ordinal action spaces. Specifically, entropy assumes uniform prior beliefs, missing the impact of a priori biases on decision-makings. KL-Divergence addresses this, however, has no notion of "nearness" of actions, and additionally, has several well known potentially undesirable properties such as the lack of symmetry, and furthermore, requires the distributions to have the same support (e.g. positive probability for all actions). Mutual information is often difficult to estimate. Here, we propose an alternative approach for modeling bounded rational RL agents utilising Wasserstein distances. This approach overcomes the aforementioned issues. Crucially, this approach accounts for the nearness of ordinal actions, modeling "stickiness" in agent decisions and unlikeliness of rapidly switching to far away actions, while also supporting low probability actions, zero-support prior distributions, and is simple to calculate directly.

It is great to see the kind of interest in applying machine learning, and specifically reinforcement learning, into real-world problems such as computer systems as presented in this paper. While the paper has no significant contributions on either a theoretical or algorithmic front, it does an important job at highlighting some of the issues in applying modern RL algorithms to real problems, and provides a necessary benchmarking environment for computer systems research specifically. The problem domains included have a wide variety of characteristics, from high-frequent real-time systems to very-long horizon problems, uniquely structured state and action spaces and both simulated and real environments (some other related work that could be added is [1]). Especially the latter is valuable to ground any research. Moreover, the authors provide an RL baseline result for each of the proposed tasks, and highlight some of the problematic characteristics of these tasks for RL specifically. There could be a more elaborate discussion of the results however.

The specification language seems to be similar to past work, being a restricted form of temporal logic. The atomic predicates comes in two flavours: ("eventually") achieve certain state or ("always") ensuring to avoid certain states. Various composition of these atomic predicates can be used (A then B, A or B, etc.). The paper's proposed finite state machine "task monitor" bears resemblance to the FSM "reward machines" proposed by Icarte et al. [1], which was not cited/discussed. So I will be quite interested how the authours clarify its differences to the Reward Machines.