We introduce PPOPT - Proximal Policy Optimization using Pretraining, a novel, model-free deep-reinforcement-learning algorithm that leverages pretraining to achieve high training efficiency and stability on very small training samples in physics-based environments. Reinforcement learning agents typically rely on large samples of environment interactions to learn a policy. However, frequent interactions with a (computer-simulated) environment may incur high computational costs, especially when the environment is complex. Our main innovation is a new policy neural network architecture that consists of a pretrained neural network middle section sandwiched between two fully-connected networks. Pretraining part of the network on a different environment with similar physics will help the agent learn the target environment with high efficiency because it will leverage a general understanding of the transferrable physics characteristics from the pretraining environment. We demonstrate that PPOPT outperforms baseline classic PPO on small training samples both in terms of rewards gained and general training stability. While PPOPT underperforms against classic model-based methods such as DYNA DDPG, the model-free nature of PPOPT allows it to train in significantly less time than its model-based counterparts. Finally, we present our implementation of PPOPT as open-source software, available at github.com/Davidrxyang/PPOPT.

The empirical success of multi-agent reinforcement learning (MARL) has motivated the search for more efficient and scalable algorithms for large scale multi-agent systems. However, existing state-of-the-art algorithms do not fully exploit inter-agent coupling information to develop MARL algorithms. In this paper, we propose a systematic approach to leverage structures in the inter-agent couplings for efficient model-free reinforcement learning. We model the cooperative MARL problem via a Bayesian network and characterize the subset of agents, termed as the value dependency set, whose information is required by each agent to estimate its local action value function exactly. Moreover, we propose a partially decentralized training decentralized execution (P-DTDE) paradigm based on the value dependency set. We theoretically establish that the total variance of our P-DTDE policy gradient estimator is less than the centralized training decentralized execution (CTDE) policy gradient estimator. We derive a multi-agent policy gradient theorem based on the P-DTDE scheme and develop a scalable actor-critic algorithm. We demonstrate the efficiency and scalability of the proposed algorithm on multi-warehouse resource allocation and multi-zone temperature control examples. For dense value dependency sets, we propose an approximation scheme based on truncation of the Bayesian network and empirically show that it achieves a faster convergence than the exact value dependence set for applications with a large number of agents.

Static feature exclusion strategies often fail to prevent bias when hidden dependencies influence the model predictions. To address this issue, we explore a reinforcement learning (RL) framework that integrates bias mitigation and automated feature selection within a single learning process. Unlike traditional heuristic-driven filter or wrapper approaches, our RL agent adaptively selects features using a reward signal that explicitly integrates predictive performance with fairness considerations. This dynamic formulation allows the model to balance generalization, accuracy, and equity throughout the training process, rather than rely exclusively on pre-processing adjustments or post hoc correction mechanisms. In this paper, we describe the construction of a multi-component reward function, the specification of the agents action space over feature subsets, and the integration of this system with ensemble learning. We aim to provide a flexible and generalizable way to select features in environments where predictors are correlated and biases can inadvertently re-emerge.

To address the dual challenges of inherent stochasticity and non-differentiable metrics in physical spatiotemporal forecasting, we propose Spatiotemporal Forecasting as Planning (SFP), a new paradigm grounded in Model-Based Reinforcement Learning. SFP constructs a novel Generative World Model to simulate diverse, high-fidelity future states, enabling an "imagination-based" environmental simulation. Within this framework, a base forecasting model acts as an agent, guided by a beam search-based planning algorithm that leverages non-differentiable domain metrics as reward signals to explore high-return future sequences. These identified high-reward candidates then serve as pseudo-labels to continuously optimize the agent's policy through iterative self-training, significantly reducing prediction error and demonstrating exceptional performance on critical domain metrics like capturing extreme events.

This paper presents a reinforcement learning (RL) based approach for path planning of cellular connected unmanned aerial vehicles (UAVs) operating beyond visual line of sight (BVLoS). The objective is to minimize travel distance while maximizing the quality of cellular link connectivity by considering real world aerial coverage constraints and employing an empirical aerial channel model. The proposed solution employs RL techniques to train an agent, using the quality of communication links between the UAV and base stations (BSs) as the reward function. Simulation results demonstrate the effectiveness of the proposed method in training the agent and generating feasible UAV path plans. The proposed approach addresses the challenges due to limitations in UAV cellular communications, highlighting the need for investigations and considerations in this area. The RL algorithm efficiently identifies optimal paths, ensuring maximum connectivity with ground BSs to ensure safe and reliable BVLoS flight operation. Moreover, the solution can be deployed as an offline path planning module that can be integrated into future ground control systems (GCS) for UAV operations, enhancing their capabilities and safety. The method holds potential for complex long range UAV applications, advancing the technology in the field of cellular connected UAV path planning.

I N T R O D U C T I O N T h e I r a n i a n l e o p a r d ( P a n t h e r a p a rd u s t u l l i a n a), a subspecies of the P ersian leopard, is critically endangered due to illegal poaching, habitat fragmentation, and h u m a n - w i l d l i f e c o n f l i c t. C o n s e r v a t i o n e f f o r t s a r e i n c r e a s i n g l y t u r n i n g t o t e c h n o l o g y f o r i n n o v a t i v e m o n i t o r i n g a n d i n t e r v e n t i o n m e t h o d s . Metric 10 Agents T raining Time (hrs) Memor y Usage (GB) CPU Utilization (%) GPU Utilization (%) T raining Time Increase (%) Memor y Usage Increase (%) 5.2 4.5 65 45 - - 20 Agents 50 Agents 6.3 5.1 75 55 20 15 8.0 6.8 85 70 53 51 T able 4. P ercentage of High-Risk zones Covered by Each Method (Mean std) F igure 3. P oacher Detection R ate Across Episodes. Higher Entropy Indicates More Diverse Exploration T able 5. KL Divergence between Inferred q(z) and Ground T ruth T ask Distribution T h e E M - b a s e d p o l i c y e x h i b i t s a n i n i t i a l l y h i g h e n t r o p y, e n c o u r a g i n g d i v e r s e a c t i o n s a m p l i n g, a n d g r a d u a l l y an n e a l s as th e po l i c y be c o m e s co n f i d e n t . Metric Cooperative Coverage Number of Agents Involved Coverage Efficiency (%) P oa ch er D et ec ti on R at e (%) Collision Incidents 6 85.3 - 0 P oacher Detection Coordination Conflict A voidance 8 - 92.1 0 10 - - 0 It enables conser vationists and security forces to allocate limited resources more effectiv e l y a n d a c t i n r e a l t i m e b a s e d o n a c t i o n a b l e i n t e l l i g e n c e d e r i v e d f r o m a u t o n o m o u s a g e n t s .

Offline reinforcement learning refers to the process of learning policies from fixed datasets, without requiring additional environment interaction. However, it often relies on well-defined reward functions, which are difficult and expensive to design. Human feedback is an appealing alternative, but its two common forms, expert demonstrations and preferences, have complementary limitations. Demonstrations provide stepwise supervision, but they are costly to collect and often reflect limited expert behavior modes. In contrast, preferences are easier to collect, but it is unclear which parts of a behavior contribute most to a trajectory segment, leaving credit assignment unresolved. In this paper, we introduce a Search-Based Preference Weighting (SPW) scheme to unify these two feedback sources. For each transition in a preference labeled trajectory, SPW searches for the most similar state-action pairs from expert demonstrations and directly derives stepwise importance weights based on their similarity scores. These weights are then used to guide standard preference learning, enabling more accurate credit assignment that traditional approaches struggle to achieve. We demonstrate that SPW enables effective joint learning from preferences and demonstrations, outperforming prior methods that leverage both feedback types on challenging robot manipulation tasks.

Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) are two prominent post-training paradigms for refining the capabilities and aligning the behavior of Large Language Models (LLMs). Existing approaches that integrate SFT and RL often face the risk of disrupting established response patterns and inducing overfitting to expert data. To address this, we present a novel investigation into the unified view of SFT and RL through an off-policy versus on-policy lens. We propose CHORD, a framework for Controllable Harmonization of On- and Off-Policy Reinforcement Learning via Dynamic Weighting, which reframes SFT not as a separate stage but as a dynamically weighted auxiliary objective within the on-policy RL process. Based on an analysis of off-policy expert data's influence at both holistic and granular levels, we incorporate a dual-control mechanism in CHORD. Specifically, the framework first employs a global coefficient to holistically guide the transition from off-policy imitation to on-policy exploration, and then applies a token-wise weighting function that enables granular learning from the expert, which promotes on-policy exploration and mitigates disruption from off-policy data. We conduct extensive experiments on mathematical reasoning problems and practical tool-use tasks, providing empirical evidence that CHORD achieves a stable and efficient learning process. By effectively harmonizing off-policy expert data with on-policy exploration, CHORD demonstrates significant improvements over baselines. We release the implementation at https://github.com/modelscope/Trinity-RFT/tree/main/examples/mix_chord to inspire further research.

Training effective Vision-Language Models (VLMs) for GUI agents typically depends on large-scale annotated datasets, whose collection is both labor-intensive and error-prone. We introduce K-step GUI Transition, a self-supervised inverse dynamics task in which VLMs learn GUI dynamics by predicting the initial action that causes a transition between two GUI states. This approach eliminates the need for natural language instructions and enables scalable dataset construction from existing GUI trajectories or automated exploration. Building on this task, we propose GUI-Shift, a reinforcement learning (RL) framework that combines rule-based optimization with data filtering to improve VLM performance. We conduct extensive experiments using multiple VLM backbones across four benchmarks, spanning GUI task automation (AndroidControl, GUI Odyssey) and GUI grounding (ScreenSpot-v2, ScreenSpot-Pro). Our results show that training on GUI-Shift generalizes well to both GUI automation and grounding tasks, yielding up to an 11.2% increase in GUI automation accuracy. This study underscores the potential of self-supervised RL to leverage unlabeled GUI trajectories and offers a scalable alternative to training with annotated samples.

Today's generative models thrive with large amounts of supervised data and informative reward functions characterizing the quality of the generation. They work under the assumptions that the supervised data provides knowledge to pre-train the model, and the reward function provides dense information about how to further improve the generation quality and correctness. However, in the hardest instances of important problems, two problems arise: (1) the base generative model attains a near-zero reward signal, and (2) calls to the reward oracle are expensive. This setting poses a fundamentally different learning challenge than standard reward-based post-training. To address this, we propose BaNEL (Bayesian Negative Evidence Learning), an algorithm that post-trains the model using failed attempts only, while minimizing the number of reward evaluations (NREs). Our method is based on the idea that the problem of learning regularities underlying failures can be cast as another, in-loop generative modeling problem. We then leverage this model to assess whether new data resembles previously seen failures and steer the generation away from them. We show that BaNEL can improve model performance without observing a single successful sample on several sparse-reward tasks, outperforming existing novelty-bonus approaches by up to several orders of magnitude in success rate, while using fewer reward evaluations.