Elastic Decision Transformers (EDTs) have proved to be particularly successful in offline reinforcement learning, offering a flexible framework that unifies sequence modeling with decision-making under uncertainty. Recent research has shown that incorporating intrinsic motivation mechanisms into EDTs improves performance across exploration tasks, yet the representational mechanisms underlying these improvements remain unexplored. In this paper, we introduce a systematic post-hoc explainability framework to analyze how intrinsic motivation shapes learned embeddings in EDTs. Through statistical analysis of embedding properties (including covariance structure, vector magnitudes, and orthogonality), we reveal that different intrinsic motivation variants create fundamentally different representational structures. Our analysis demonstrates environment-specific correlation patterns between embedding metrics and performance that explain why intrinsic motivation improves policy learning. These findings show that intrinsic motivation operates beyond simple exploration bonuses, acting as a representational prior that shapes embedding geometry in biologically plausible ways, creating environment-specific organizational structures that facilitate better decision-making.

Recent advances have demonstrated that integrating reinforcement learning with rule-based rewards can significantly enhance the reasoning capabilities of large language models, even without supervised fine-tuning. However, prevalent reinforcement learning algorithms such as GRPO and its variants like DAPO, suffer from a coarse granularity issue when computing the advantage. Specifically, they compute rollout-level advantages that assign identical values to every token within a sequence, failing to capture token-specific contributions and hindering effective learning. To address this limitation, we propose Key-token Advantage Estimation (KTAE) - a novel algorithm that estimates fine-grained, token-level advantages without introducing additional models. KTAE leverages the correctness of sampled rollouts and applies statistical analysis to quantify the importance of individual tokens within a sequence to the final outcome. This quantified token-level importance is then combined with the rollout-level advantage to obtain a more fine-grained token-level advantage estimation. Empirical results show that models trained with GRPO+KTAE and DAPO+KTAE outperform baseline methods across five mathematical reasoning benchmarks. Notably, they achieve higher accuracy with shorter responses and even surpass R1-Distill-Qwen-1.5B using the same base model.

Inspired by foveal vision, hard attention models promise interpretability and parameter economy. However, existing models like the Recurrent Model of Visual Attention (RAM) and Deep Recurrent Attention Model (DRAM) failed to model the hierarchy of human vision system, that compromise on the visual exploration dynamics. As a result, they tend to produce attention that are either overly fixational or excessively saccadic, diverging from human eye movement behavior. In this paper, we propose a Multi-Level Recurrent Attention Model (MRAM), a novel hard attention framework that explicitly models the neural hierarchy of human visual processing. By decoupling the function of glimpse location generation and task execution in two recurrent layers, MRAM emergent a balanced behavior between fixation and saccadic movement. Our results show that MRAM not only achieves more human-like attention dynamics, but also consistently outperforms CNN, RAM and DRAM baselines on standard image classification benchmarks.

Humanoid loco-manipulation holds transformative potential for daily service and industrial tasks, yet achieving precise, robust whole-body control with 3D end-effector force interaction remains a major challenge. Prior approaches are often limited to lightweight tasks or quadrupedal/wheeled platforms. To overcome these limitations, we propose FALCON, a dual-agent reinforcement-learning-based framework for robust force-adaptive humanoid loco-manipulation. FALCON decomposes whole-body control into two specialized agents: (1) a lower-body agent ensuring stable locomotion under external force disturbances, and (2) an upper-body agent precisely tracking end-effector positions with implicit adaptive force compensation. These two agents are jointly trained in simulation with a force curriculum that progressively escalates the magnitude of external force exerted on the end effector while respecting torque limits. Experiments demonstrate that, compared to the baselines, FALCON achieves 2x more accurate upper-body joint tracking, while maintaining robust locomotion under force disturbances and achieving faster training convergence. Moreover, FALCON enables policy training without embodiment-specific reward or curriculum tuning. Using the same training setup, we obtain policies that are deployed across multiple humanoids, enabling forceful loco-manipulation tasks such as transporting payloads (0-20N force), cart-pulling (0-100N), and door-opening (0-40N) in the real world.

This paper presents Post-Decision Proximal Policy Optimization (PDPPO), a novel variation of the leading deep reinforcement learning method, Proximal Policy Optimization (PPO). The PDPPO state transition process is divided into two steps: a deterministic step resulting in the post-decision state and a stochastic step leading to the next state. Our approach incorporates post-decision states and dual critics to reduce the problem's dimensionality and enhance the accuracy of value function estimation. Lot-sizing is a mixed integer programming problem for which we exemplify such dynamics. The objective of lot-sizing is to optimize production, delivery fulfillment, and inventory levels in uncertain demand and cost parameters. This paper evaluates the performance of PDPPO across various environments and configurations. Notably, PDPPO with a dual critic architecture achieves nearly double the maximum reward of vanilla PPO in specific scenarios, requiring fewer episode iterations and demonstrating faster and more consistent learning across different initializations. On average, PDPPO outperforms PPO in environments with a stochastic component in the state transition. These results support the benefits of using a post-decision state. Integrating this post-decision state in the value function approximation leads to more informed and efficient learning in high-dimensional and stochastic environments.

Long-horizon planning is crucial in complex environments, but diffusion-based planners like Diffuser are limited by the trajectory lengths observed during training. This creates a dilemma: long trajectories are needed for effective planning, yet they degrade model performance. In this paper, we introduce this extendable long-horizon planning challenge and propose a two-phase solution. First, Progressive Trajectory Extension incrementally constructs longer trajectories through multi-round compositional stitching. Second, the Hierarchical Multiscale Diffuser enables efficient training and inference over long horizons by reasoning across temporal scales. To avoid the need for multiple separate models, we propose Adaptive Plan Pondering and the Recursive HM-Diffuser, which unify hierarchical planning within a single model. Experiments show our approach yields strong performance gains, advancing scalable and efficient decision-making over long-horizons.

Abstract--Reinforcement learning (RL) has made significant advancements, achieving superhuman performance in various tasks. However, RL agents often operate under the assumption of environmental stationarity, which poses a great challenge to learning efficiency since many environments are inherently non-stationary. T o address this issue, we introduce the Clustering Orthogonal Weight Modified (COWM) layer, which can be integrated into the policy network of any RL algorithm and mitigate non-stationarity effectively. The COWM layer stabilizes the learning process by employing clustering techniques and a projection matrix. Our approach not only improves learning speed but also reduces gradient interference, thereby enhancing the overall learning efficiency. Empirically, the COWM outperforms state-of-the-art methods and achieves improvements of 9% and 12.6% in vision-based and state-based DMControl benchmark. It also shows robustness and generality across various algorithms and tasks. In recent years, reinforcement learning (RL) has made significant progress across various domains, ranging from gaming to robotic control, often surpassing human performance [1]- [6]. Despite these advancements, a significant issue remains: the underlying assumption of a stationary environment [7].

Cloud resource allocation has emerged as a major challenge in modern computing environments, with organizations struggling to manage complex, dynamic workloads while optimizing performance and cost efficiency. Traditional heuristic approaches prove inadequate for handling the multi-objective optimization demands of existing cloud infrastructures. This paper presents a comparative analysis of state-of-the-art artificial intelligence and machine learning algorithms for resource allocation. We systematically evaluate 10 algorithms across four categories: Deep Reinforcement Learning approaches, Neural Network architectures, Traditional Machine Learning enhanced methods, and Multi-Agent systems. Analysis of published results demonstrates significant performance improvements across multiple metrics including makespan reduction, cost optimization, and energy efficiency gains compared to traditional methods. The findings reveal that hybrid architectures combining multiple artificial intelligence and machine learning techniques consistently outperform single-method approaches, with edge computing environments showing the highest deployment readiness. Our analysis provides critical insights for both academic researchers and industry practitioners seeking to implement next-generation cloud resource allocation strategies in increasingly complex and dynamic computing environments.

Most estimators collapse all uncertainty modes into a single confidence score, preventing reliable reasoning about when to allocate more compute or adjust inference. W e introduce Uncertainty-Guided Inference-Time Selection, a lightweight inference time framework that disentangles aleatoric (data-driven) and epistemic (model-driven) uncertainty directly in deep feature space. Aleatoric uncertainty is estimated using a regularized global density model, while epistemic uncertainty is formed from three complementary components that capture local support deficiency, manifold spectral collapse, and cross-layer feature inconsistency. These components are empirically orthogonal and require no sampling, no ensembling, and no additional forward passes. W e integrate the decomposed uncertainty into a distribution free conformal calibration procedure that yields significantly tighter prediction intervals at matched coverage. Using these components for uncertainty guided adaptive model selection reduces compute by approximately 60 percent on MOT17 with negligible accuracy loss, enabling practical self regulating visual inference. Additionally, our ablation results show that the proposed orthogonal uncertainty decomposition consistently yields higher computational savings across all MOT17 sequences, improving margins by 13.6 percentage points over the total-uncertainty baseline.

Stock trading strategies play a critical role in investment. However, it is challenging to design a profitable strategy in a complex and dynamic stock market. In this paper, we propose an ensemble strategy that employs deep reinforcement schemes to learn a stock trading strategy by maximizing investment return. We train a deep reinforcement learning agent and obtain an ensemble trading strategy using three actor-critic based algorithms: Proximal Policy Optimization (PPO), Advantage Actor Critic (A2C), and Deep Deterministic Policy Gradient (DDPG). The ensemble strategy inherits and integrates the best features of the three algorithms, thereby robustly adjusting to different market situations. In order to avoid the large memory consumption in training networks with continuous action space, we employ a load-on-demand technique for processing very large data. We test our algorithms on the 30 Dow Jones stocks that have adequate liquidity. The performance of the trading agent with different reinforcement learning algorithms is evaluated and compared with both the Dow Jones Industrial Average index and the traditional min-variance portfolio allocation strategy. The proposed deep ensemble strategy is shown to outperform the three individual algorithms and two baselines in terms of the risk-adjusted return measured by the Sharpe ratio. This work is fully open-sourced at \href{https://github.com/AI4Finance-Foundation/Deep-Reinforcement-Learning-for-Automated-Stock-Trading-Ensemble-Strategy-ICAIF-2020}{GitHub}.