Transformer-Based RL

In dynamic urban logistics, the stochastic emergence of time-sensitive tasks poses a significant optimality challenge for heterogeneous AAVs logistics task allocation. To address this problem, a reinforcement learning enhanced overlapping coalition formation game approach is proposed. A dynamic task allocation model is established, where global optimality is mathematically quantified by a generalized logistics cost coupling service quality and resource consumption. To deal with the time-varying task sets induced by stochastic order arrivals, a transformer-based soft actor-critic network is designed. By leveraging multi-head self-attention to encode variable-length logistics states and capture task-wise spatiotemporal dependencies, the learned policy adaptively guides coalition updates, replacing heuristic rules in the overlapping coalition formation game. On this basis, heterogeneous AAVs can form more efficient overlapping coalitions for dynamic logistics tasks. The resulting coalition formation process is proven to constitute an exact potential game, which guarantees convergence to a Nash-stable equilibrium within a finite number of iterations. Numerical simulations demonstrate that the proposed algorithm effectively improves the optimality of task allocation under the generalized logistics cost criterion. In a scenario with 32 AAVs and 80 tasks, our algorithm achieves a 39.76% cost reduction compared with the heuristic OCF baseline. Indoor flight experiments further validate its practicality.

Token-based transformer world models have shown strong performance in visual reinforcement learning, but often suffer from temporal inconsistency in long-horizon rollouts, including object duplication, disappearance, and transmutation. A key reason is that most existing approaches treat next-frame prediction purely as a token generation problem, without considering the persistence of tokens across time. We introduce Identifiable Token Correspondence (ITC), a decoding step for token-based transformer world models that formulates next-frame prediction as a structured assignment problem with latent token correspondence variables: each next-frame token is explained either by copying a token from the previous frame or by generating a new one. ITC leaves the transformer architecture and training procedure unchanged and can be added on top of existing backbones. Our experiments show state-of-the-art performance on 4 challenging benchmarks. The proposed method achieves a return of 72.5% and a score of 35.6% on the Craftax-classic benchmark, significantly surpassing the previous best of 67.4% and 27.9%. We release our source code on https://github.com/snu-mllab/Identifiable-Token-Correspondence.

A central challenge in reinforcement learning (RL) is to learn models that generalize beyond the tasks on which they are trained, a goal traditionally pursued through multi-task and meta RL. Recently, transformer architectures have emerged as a promising approach, enabling adaptation to new tasks via in-context learning without explicit parameter updates. From a functional perspective, a transformer can be viewed as a functional operator that maps a context to a task-specific function. It is thus fundamental to understand and design this operator to support stronger generalization in RL. In this work, we address this resulting question of generalization from a kernel-based perspective by establishing a connection between non-linear transformers and kernel-based temporal difference learning. By interpreting the transformer as performing regression in a Reproducing Kernel Hilbert Space (RKHS), we show that value functions from different domains can be represented using a shared set of weights, provided they lie within the same RKHS. Experiments on multiple MetaWorld domains support this interpretation, demonstrating convergence of the temporal-difference objective.

Long-horizon, sparse-reward tasks pose a fundamental challenge for reinforcement learning, since single-step TD learning suffers from bootstrapping error accumulation across successive Bellman updates. Actor-critic methods with action chunking address this by operating over temporally extended actions, which reduce the effective horizon, enable fast value backups, and support temporally consistent exploration. However, existing methods rely on a fixed chunk size and therefore cannot adaptively balance reactivity against temporal consistency. A large fixed chunk size reduces responsiveness to new observations, while a small one produces incoherent motions, forcing task-specific tuning of the chunk size. To address this limitation, we propose Adaptive Chunk Size Actor-Critic (ACSAC). ACSAC leverages a causal Transformer critic to evaluate expected returns for action chunks of different sizes. At each chunk boundary, it adaptively selects the chunk size that maximizes the expected return, supporting flexible, state-dependent chunk sizes without task-specific tuning. We prove that the ACSAC Bellman operator is a contraction whose unique fixed point is the action-value function of the adaptive policy. Experiments on OGBench demonstrate that ACSAC achieves state-of-the-art performance on long-horizon, sparse-reward manipulation tasks across both offline RL and offline-to-online RL settings.

In-context reinforcement learning (ICRL) studies agents that, after pretraining, adapt to new tasks by conditioning on additional context without parameter updates. Existing theoretical analyses of ICRL largely rely on linear attention, which replaces the softmax function in the standard attention with an identity mapping. This paper provides the first theoretical understanding of ICRL without making the unrealistic linear attention simplification. In particular, we consider the standard softmax attention used in practice. We show that, with certain parameters, the layerwise forward pass of a Transformer with such softmax attention is equivalent to iterative updates of a weighted softmax temporal difference (TD) learning algorithm. Here, weighted softmax TD is a new RL algorithm that performs policy evaluation in kernel space and adopts both linear TD and tabular TD as special cases. We also prove that under a certain contraction condition, the policy evaluation error decays as the number of layers grows, with the identified parameters above. Finally, we prove that those parameters are a global minimizer of a pretraining loss, explaining their emergence in our numerical experiments.

In-context reinforcement learning (ICRL) refers to the ability of RL agents to adapt to new tasks at inference time without parameter updates by conditioning on additional context. Recent empirical studies further demonstrate that Chain-of-Thought (CoT) generation can amplify this ICRL capability. This paper is the first to provide a theoretical understanding on how CoT interacts with ICRL. We conduct our analysis in a policy evaluation setup with linear Transformer. We prove that with specific Transformer parameters, the CoT generation process is equivalent to repeatedly executing temporal difference learning updates. Additionally, we provide finite sample convergence analysis showing that the policy evaluation error decreases geometrically with CoT length and eventually saturates at a statistical floor determined by the context length. We also prove that the desired Transformer parameters are a global minimizer of the pretraining loss, providing a theoretical understanding on the empirical emergence of those parameters.

Decision Transformer (DT) formulates offline reinforcement learning as autoregressive sequence modeling, achieving promising results by predicting actions from a sequence of Return-to-Go (RTG), state, and action tokens. However, RTG is a scalar that summarizes future rewards, containing far less information than typical state or action vectors, yet it consumes the same computational budget per token. Worse, the self-attention cost of Transformers grows quadratically with sequence length, so including RTG as a separate token adds unnecessary overhead. We propose SlimDT, which removes RTG from the autoregressive sequence. Instead, we inject RTG information into the state representations before the sequential modeling step, allowing the Transformer to process only a compact (state, action) sequence. This reduces the sequence length by one-third, directly improving inference efficiency. On the D4RL benchmark, SlimDT surpasses standard DT across various tasks and achieves performance comparable to existing state-of-the-art methods. Decoupling a sparse conditioning signal from an information-rich sequence thus yields both computational gains and higher task performance.

We investigate the ability of transformers to perform in-context reinforcement learning (ICRL), where a model must infer and execute learning algorithms from trajectory data without parameter updates. We show that a linear self-attention transformer block can provably implement policy-improvement methods, including semi-gradient SARSA and actor-critic, via explicit parameter constructions. Beyond existence, we design a teacher-mimicking training procedure, analyze its gradient-flow dynamics, and establish the first convergence guarantee in the ICRL literature: under suitable richness conditions on the training MDP distribution, gradient flow converges locally and exponentially to an optimal parameter manifold corresponding to the desired RL update. Empirically, training transformers on randomly generated tabular MDPs confirms these predictions: the learned models recover the parameter structure of our explicit constructions and, when deployed on unseen MDPs, deliver strong in-context control performance. Together, these results illuminate how transformer architectures internalize and execute classical reinforcement learning algorithms in context, bridging mechanistic understanding and training dynamics in ICRL.

Cooperative multi-agent reinforcement learning (MARL) is widely used to address large joint observation and action spaces by decomposing a centralized control problem into multiple interacting agents. However, such decomposition often introduces additional challenges, including non-stationarity, unstable training, weak coordination, and limited theoretical guarantees. In this paper, we propose the Consensus Multi-Agent Transformer (CMAT), a centralized framework that bridges cooperative MARL to a hierarchical single-agent reinforcement learning (SARL) formulation. CMAT treats all agents as a unified entity and employs a Transformer encoder to process the large joint observation space. To handle the extensive joint action space, we introduce a hierarchical decision-making mechanism in which a Transformer decoder autoregressively generates a high-level consensus vector, simulating the process by which agents reach agreement on their strategies in latent space. Conditioned on this consensus, all agents generate their actions simultaneously, enabling order-independent joint decision making and avoiding the sensitivity to action-generation order in conventional Multi-Agent Transformers (MAT). This factorization allows the joint policy to be optimized using single-agent PPO while preserving expressive coordination through the latent consensus. To evaluate the proposed method, we conduct experiments on benchmark tasks from StarCraft II, Multi-Agent MuJoCo, and Google Research Football. The results show that CMAT achieves superior performance over recent centralized solutions, sequential MARL methods, and conventional MARL baselines. The code for this paper is available at:https://github.com/RS2002/CMAT .

Recent advances in multi-agent reinforcement learning (MARL) have demonstrated success in numerous challenging domains and environments, but typically require specialized models for each task. In this work, we propose a coherent methodology that makes it possible for a single GPT-based model to learn and perform well across diverse MARL environments and tasks, including StarCraft Multi-Agent Challenge, Google Research Football and POGEMA. Our method, MARL-GPT, applies offline reinforcement learning to train at scale on the expert trajectories (400M for SMACv2, 100M for GRF, and 1B for POGEMA) combined with a single transformer-based observation encoder that requires no task-specific tuning. Experiments show that MARL-GPT achieves competitive performance compared to specialized baselines in all tested environments. Thus, our findings suggest that it is, indeed, possible to build a multi-task transformer-based model for a wide variety of (significantly different) multi-agent problems paving the way to the fundamental MARL model (akin to ChatGPT, Llama, Mistral etc. in natural language modeling).

Recent progress in in-context reinforcement learning (ICRL) has demonstrated its potential for training generalist agents that can acquire new tasks directly at inference. Algorithm Distillation (AD) pioneered this paradigm and was subsequently scaled to multi-domain settings, although its ability to generalize to unseen tasks remained limited. The Decision Pre-Trained Transformer (DPT) was introduced as an alternative, showing stronger in-context reinforcement learning abilities in simplified domains, but its scalability had not been established. In this work, we extend DPT to diverse multi-domain environments, applying Flow Matching as a natural training choice that preserves its interpretation as Bayesian posterior sampling. As a result, we obtain an agent trained across hundreds of diverse tasks that achieves clear gains in generalization to the held-out test set. This agent improves upon prior AD scaling and demonstrates stronger performance in both online and offline inference, reinforcing ICRL as a viable alternative to expert distillation for training generalist agents.

Drones equipped with overhead manipulators offer unique capabilities for inspection, maintenance, and contact-based interaction. However, the motion of the drone and its manipulator is tightly linked, and even small attitude changes caused by wind or control imperfections shift the end-effector away from its intended path. This coupling makes reliable tracking difficult and also limits the direct use of learning-based arm controllers that were originally designed for fixed-base robots. These effects appear consistently in our tests whenever the UAV body experiences drift or rapid attitude corrections. To address this behavior, we develop a reinforcement-learning (RL) framework with a transformer-based double deep Q learning (DDQN), with the core idea of using an adaptive beam-search planner that applies a short-horizon beam search over candidate control sequences using the learned critic as the forward estimator. This allows the controller to anticipate the end-effector's motion through simulated rollouts rather than executing those actions directly on the actual model, realizing a software-in-the-loop (SITL) approach. The lookahead relies on value estimates from a Transformer critic that processes short sequences of states, while a DDQN backbone provides the one-step targets needed to keep the learning process stable. Evaluated on a 3-DoF aerial manipulator under identical training conditions, the proposed meta-adaptive planner shows the strongest overall performance with a 10.2% reward increase, a substantial reduction in mean tracking error (from about 6% to 3%), and a 29.6% improvement in the combined reward-error metric relative to the DDQN baseline. Our method exhibits elevated stability in tracking target tip trajectory (by maintaining 5 cm tracking error) when the drone base exhibits drifts due to external disturbances, as opposed to the fixed-beam and Transformer-only variants.

Transformer-based reinforcement learning has emerged as a strong candidate for sequential control in residential energy management. In particular, the Decision Transformer can learn effective battery dispatch policies from historical data, thereby increasing photovoltaic self-consumption and reducing electricity costs. However, transformer models are typically too computationally demanding for deployment on resource-constrained residential controllers, where memory and latency constraints are critical. This paper investigates knowledge distillation to transfer the decision-making behaviour of high-capacity Decision Transformer policies to compact models that are more suitable for embedded deployment. Using the Ausgrid dataset, we train teacher models in an offline sequence-based Decision Transformer framework on heterogeneous multi-building data. We then distil smaller student models by matching the teachers' actions, thereby preserving control quality while reducing model size. Across a broad set of teacher-student configurations, distillation largely preserves control performance and even yields small improvements of up to 1%, while reducing the parameter count by up to 96%, the inference memory by up to 90%, and the inference time by up to 63%. Beyond these compression effects, comparable cost improvements are also observed when distilling into a student model of identical architectural capacity. Overall, our results show that knowledge distillation makes Decision Transformer control more applicable for residential energy management on resource-limited hardware.

The open shop scheduling problem (OSSP) arises in many industrial and service settings but remains computationally challenging as the number of jobs and machines increases. While exact methods quickly become intractable, classical dispatching rules and metaheuristics may require substantial tuning to maintain solution quality at large scales. This study develops a Transformer-based scheduling policy for OSSP using an encoder-decoder architecture with multi-head attention. The model is trained on Taillard benchmark instances (4x4, 5x5, 7x7, and 10x10) using only the processing-time matrix as input and produces feasible schedules with makespans typically within 15-30% of best-known values. To evaluate scalability, the trained policy is applied without retraining to randomly generated instances from 40x40 to 100x100 and compared against classical dispatching heuristics, including SPT, LPT, MWKR, and EST. Across these large instances, the Transformer achieved average gaps of 12.89-15.12% relative to a standard lower bound. Compared with EST, the Transformer remained competitive, typically within a modest margin, while substantially outperforming SPT and LPT. These results indicate that a Transformer policy trained on small OSSP instances can generalize to substantially larger problems and provide a feature-light, learning-based alternative to classical dispatching rules.

Offline multi-agent reinforcement learning (MARL) with multi-task datasets is challenging due to varying numbers of agents across tasks and the need to generalize to unseen scenarios. Prior works employ transformers with observation tokenization and hierarchical skill learning to address these issues. However, they underutilize the transformer attention mechanism for inter-agent coordination and rely on a single history token, which limits their ability to capture long-horizon temporal dependencies in partially observable MARL settings. In this paper, we propose STAIRS-Former, a transformer architecture augmented with spatial and temporal hierarchies that enables effective attention over critical tokens while capturing long interaction histories. We further introduce token dropout to enhance robustness and generalization across varying agent populations. Extensive experiments on diverse multi-agent benchmarks, including SMAC, SMAC-v2, MPE, and MaMuJoCo, with multi-task datasets demonstrate that STAIRS-Former consistently outperforms prior methods and achieves new state-of-the-art performance.

Mechanical ventilation (MV) is a life-saving intervention for patients with acute respiratory failure (ARF) in the ICU. However, inappropriate ventilator settings could cause ventilator-induced lung injury (VILI). Also, clinicians workload is shown to be directly linked to patient outcomes. Hence, MV should be personalized and automated to improve patient outcomes. Previous attempts to incorporate personalization and automation in MV include traditional supervised learning and offline reinforcement learning (RL) approaches, which often neglect temporal dependencies and rely excessively on mortality-based rewards. As a result, early stage physiological deterioration and the risk of VILI are not adequately captured. To address these limitations, we propose Transformer-based Conservative Q-Learning (T-CQL), a novel offline RL framework that integrates a Transformer encoder for effective temporal modeling of patient dynamics, conservative adaptive regularization based on uncertainty quantification to ensure safety, and consistency regularization for robust decision-making. We build a clinically informed reward function that incorporates indicators of VILI and a score for severity of patients illness. Also, previous work predominantly uses Fitted Q-Evaluation (FQE) for RL policy evaluation on static offline data, which is less responsive to dynamic environmental changes and susceptible to distribution shifts. To overcome these evaluation limitations, interactive digital twins of ARF patients were used for online "at the bedside" evaluation. Our results demonstrate that T-CQL consistently outperforms existing state-of-the-art offline RL methodologies, providing safer and more effective ventilatory adjustments. Our framework demonstrates the potential of Transformer-based models combined with conservative RL strategies as a decision support tool in critical care.

Neural approaches to the Flexible Job Shop Scheduling Problem (FJSP), particularly those based on deep reinforcement learning (DRL), have gained growing attention in recent years. However, existing methods rely on complex feature-engineered state representations (i.e., often requiring more than 20 handcrafted features) and graph-biased neural architectures. To reduce modeling complexity and advance a more generalizable framework for FJSP, we introduce ReSched, a minimalist DRL framework that rethinks both the scheduling formulation and model design. First, by revisiting the Markov Decision Process (MDP) formulation of FJSP, we condense the state space to just four essential features, eliminating historical dependencies through a subproblem-based perspective. Second, we employ Transformer blocks with dot-product attention, augmented by three lightweight but effective architectural modifications tailored to scheduling tasks. Extensive experiments show that ReSched outperforms classical dispatching rules and state-of-the-art DRL methods on FJSP. Moreover, ReSched also generalizes well to the Job Shop Scheduling Problem (JSSP) and the Flexible Flow Shop Scheduling Problem (FFSP), achieving competitive performance against neural baselines specifically designed for these variants.

Real-world reinforcement learning systems must operate under distributional drift in their observation streams, yet most policy architectures implicitly assume fully observed and noise-free states. We study robustness of Proximal Policy Optimization (PPO) under temporally persistent sensor failures that induce partial observability and representation shift. To respond to this drift, we augment PPO with temporal sequence models, including Transformers and State Space Models (SSMs), to enable policies to infer missing information from history and maintain performance. Under a stochastic sensor failure process, we prove a high-probability bound on infinite-horizon reward degradation that quantifies how robustness depends on policy smoothness and failure persistence. Empirically, on MuJoCo continuous-control benchmarks with severe sensor dropout, we show Transformer-based sequence policies substantially outperform MLP, RNN, and SSM baselines in robustness, maintaining high returns even when large fractions of sensors are unavailable. These results demonstrate that temporal sequence reasoning provides a principled and practical mechanism for reliable operation under observation drift caused by sensor unreliability.

Decision transformer based sequential policies have emerged as a powerful paradigm in offline reinforcement learning (RL), yet their efficacy remains constrained by the quality of static datasets and inherent architectural limitations. Specifically, these models often struggle to effectively integrate suboptimal experiences and fail to explicitly plan for an optimal policy. To bridge this gap, we propose Imaginary Planning Distillation (IPD), a novel framework that seamlessly incorporates offline planning into data generation, supervised training, and online inference. Our framework first learns a world model equipped with uncertainty measures and a quasi-optimal value function from the offline data. These components are utilized to identify suboptimal trajectories and augment them with reliable, imagined optimal rollouts generated via Model Predictive Control (MPC). A Transformer-based sequential policy is then trained on this enriched dataset, complemented by a value-guided objective that promotes the distillation of the optimal policy. By replacing the conventional, manually-tuned return-to-go with the learned quasi-optimal value function, IPD improves both decision-making stability and performance during inference. Empirical evaluations on the D4RL benchmark demonstrate that IPD significantly outperforms several state-of-the-art value-based and transformer-based offline RL methods across diverse tasks.

Capturing temporal dependencies is critical for model-based reinforcement learning (MBRL) in partially observable, high-dimensional domains. We introduce NE-Dreamer, a decoder-free MBRL agent that leverages a temporal transformer to predict next-step encoder embeddings from latent state sequences, directly optimizing temporal predictive alignment in representation space. This approach enables NE-Dreamer to learn coherent, predictive state representations without reconstruction losses or auxiliary supervision. On the DeepMind Control Suite, NE-Dreamer matches or exceeds the performance of DreamerV3 and leading decoder-free agents. On a challenging subset of DMLab tasks involving memory and spatial reasoning, NE-Dreamer achieves substantial gains. These results establish next-embedding prediction with temporal transformers as an effective, scalable framework for MBRL in complex, partially observable environments.

Generative models have gained significant traction in offline reinforcement learning (RL) due to their ability to model complex trajectory distributions. However, existing generation-based approaches still struggle with long-horizon tasks characterized by sparse rewards. Some hierarchical generation methods have been developed to mitigate this issue by decomposing the original problem into shorter-horizon subproblems using one policy and generating detailed actions with another. While effective, these methods often overlook the multi-scale temporal structure inherent in trajectories, resulting in suboptimal performance. To overcome these limitations, we propose MAGE, a Multi-scale Autoregressive GEneration-based offline RL method. MAGE incorporates a condition-guided multi-scale autoencoder to learn hierarchical trajectory representations, along with a multi-scale transformer that autoregressively generates trajectory representations from coarse to fine temporal scales. MAGE effectively captures temporal dependencies of trajectories at multiple resolutions. Additionally, a condition-guided decoder is employed to exert precise control over short-term behaviors. Extensive experiments on five offline RL benchmarks against fifteen baseline algorithms show that MAGE successfully integrates multi-scale trajectory modeling with conditional guidance, generating coherent and controllable trajectories in long-horizon sparse-reward settings.

Real-time schedulers must reason about tight deadlines under strict compute budgets. We present TempoNet, a reinforcement learning scheduler that pairs a permutation-invariant Transformer with a deep Q-approximation. An Urgency Tokenizer discretizes temporal slack into learnable embeddings, stabilizing value learning and capturing deadline proximity. A latency-aware sparse attention stack with blockwise top-k selection and locality-sensitive chunking enables global reasoning over unordered task sets with near-linear scaling and sub-millisecond inference. A multicore mapping layer converts contextualized Q-scores into processor assignments through masked-greedy selection or differentiable matching. Extensive evaluations on industrial mixed-criticality traces and large multiprocessor settings show consistent gains in deadline fulfillment over analytic schedulers and neural baselines, together with improved optimization stability. Diagnostics include sensitivity analyses for slack quantization, attention-driven policy interpretation, hardware-in-the-loop and kernel micro-benchmarks, and robustness under stress with simple runtime mitigations; we also report sample-efficiency benefits from behavioral-cloning pretraining and compatibility with an actor-critic variant without altering the inference pipeline. These results establish a practical framework for Transformer-based decision making in high-throughput real-time scheduling.

We investigate how embedding dimension affects the emergence of an internal "world model" in a transformer trained with reinforcement learning to perform bubble-sort-style adjacent swaps. Models achieve high accuracy even with very small embedding dimensions, but larger dimensions yield more faithful, consistent, and robust internal representations. In particular, higher embedding dimensions strengthen the formation of structured internal representation and lead to better interpretability. After hundreds of experiments, we observe two consistent mechanisms: (1) the last row of the attention weight matrix monotonically encodes the global ordering of tokens; and (2) the selected transposition aligns with the largest adjacent difference of these encoded values. Our results provide quantitative evidence that transformers build structured internal world models and that model size improves representation quality in addition to end performance. We release our metrics and analyses, which can be used to probe similar algorithmic tasks.