Diffusion policy sampling enables reinforcement learning (RL) to represent multimodal action distributions beyond suboptimal unimodal Gaussian policies. However, existing diffusion-based RL methods primarily focus on offline setting for reward maximization, with limited consideration of safety in online settings. To address this gap, we propose Augmented Lagrangian-Guided Diffusion (ALGD), a novel algorithm for off-policy safe RL. By revisiting optimization theory and energy-based modeling, we show that the instability of primal–dual methods arises from the non-convex Lagrangian landscape. In diffusion-based safe RL, the Lagrangian can be interpreted as an energy function guiding the denoising dynamics; counter-intuitively, direct usage destabilizes both policy generation and training. ALGD resolves this issue by introducing an augmented Lagrangian that locally convexifies the energy landscape, yielding a stabilized policy generation and training, without altering the distribution of optimal policy. Theoretical analysis and extensive experiments demonstrate that ALGD is both theoretically grounded and empirically effective, achieving strong and stable performance across diverse environments.

Reliable value estimation serves as the cornerstone of reinforcement learning (RL) by evaluating long-term returns and guiding policy improvement, significantly influencing the convergence speed and final performance. Existing works improve the reliability of value function estimation via multi-critic ensembles and distributional RL, yet the former merely combines multi point estimation without capturing distributional information, whereas the latter relies on discretization or quantile regression, limiting the expressiveness of complex value distributions. Inspired by flow matching's success in generative modeling, we propose a generative paradigm for value estimation, named FlowCritic. Departing from conventional regression for deterministic value prediction, Flow Critic leverages flow matching to model value distributions and generate samples for value estimation. Specifically, FlowCritic formulates the target value distribution through distributional Bellman operators, and trains a velocity field network to learn the continuous probability flow from a tractable prior distribution to the target distribution, thereby enabling flexible modeling of arbitrarily complex value distributions. More importantly, FlowCritic introduces the coefficient of variation (CoV) of the generated value distribution to quantify the noise level of training samples. BasedonCoV, FlowCritic assigns adaptive weights to prioritize low noise samples, thereby reducing the variance of policy gradient during backpropagation. Additionally, FlowCritic incorporates truncated sampling and velocity field clipping mechanisms to ensure training stability. Our theoretical analysis demonstrates FlowCritic's convergence and advantages. Extensive experiments on 12 IsaacGym benchmarks demonstrate FlowCritic's superiority over existing RL baselines, while its successful deployment on a real quadrupedal robot platform validates its effectiveness in practical physical systems. Notably, FlowCritic is the first approach to integrate flow matching into value distribution modeling for RL, and offers a new perspective on the effective exploitation of distributional information.

Conversational agents are required to respond to their users not only with high quality (i.e. commonsense bearing) responses, but also considering multiple plausible alternative scenarios, reflecting the diversity in their responses. Despite the growing need to train diverse commonsense generators, the progress of this line of work has been significantly hindered by the lack of large-scale high-quality diverse commonsense training datasets. Due to the high annotation costs, existing Generative Commonsense Reasoning (GCR) datasets are created using a small number of human annotators, covering only a narrow set of commonsense scenarios. To address this training resource gap, we propose a two-stage method to create the first-ever synthetic dataset CommonSyn for diversified (GCR). The model fine-tuned on our synthetic data jointly increase both generation diversity and quality compared with vanilla models and the model fine-tuned on human-crafted dataset across different size Large Language Models (LLMs)

Distributional reinforcement learning enhances policy robustness and learning efficiency by modeling the full value distribution rather than merely estimating its expectation. However, most existing methods are designed for single-agent discrete tasks and typically rely on prior assumptions about the form of the value distribution, such as fixed discrete atoms or quantile regression. These assumptions become restrictive in multi-agent scenarios, where information from different agents must be fused and the resulting joint value distribution often exhibits complex multimodal characteristics that cannot be effectively captured by predefined distributional forms. To address the above limitations, this paper proposes Multi-Agent Deep Denoising Diffusion Policy Gradient Optimization (MAD3PG), a novel framework that leverages diffusion models to model the value distribution of joint state-action pairs. Unlike existing distributional methods that rely on prior assumptions, diffusion models inherently possess the capacity to capture value distribution, thereby demonstrating clear advantages in multi-agent environments. Since true value samples are unavailable during online training, we introduce a K-repeated sampling strategy combined with temporal-difference targets to efficiently optimize the conditional variational evidence lower bound. Furthermore, we provide a theoretical convergence analysis of MAD3PG. We demonstrate MAD3PG’s ability to fit multi-modal distributions through toy examples. Furthermore, we validate its effectiveness by conducting comprehensive experiments in the Multi-Agent Particle Environment (MPE) as well as in the MuJoCo environment. The experimental results indicate that MAD3PG significantly outperforms traditional algorithms based on deterministic value estimation. The code is available at https://github.com/smartXiaoz/MAD3PG.

Large language models (LLMs) demonstrate strong performance on standard content moderation benchmarks. However, these benchmarks often aggregate multiple moderation criteria into a single label, making it unclear whether models can disentangle them and reliably apply each criterion when making decisions. To study whether LLMs exhibit criterion-conditioned behaviour, we introduce Diagnostic Evaluation of COntent (DECO), a criterion-independent factorisation of content that enables controlled, criterion-level evaluation. We also introduce pairwise evaluation to compare model outputs across different criteria for the same input. Across four moderation datasets and four LLMs, we find that strong benchmark performance can hide substantial failures at the criterion level. Models struggle most when correct decisions depend not on overall harmfulness, but on the specific aspect of the content that the criterion requires them to assess. Our results highlight a key limitation of current content moderation benchmarks: strong performance on aggregated labels does not provide sufficient evidence that LLMs can reliably evaluate content with respect to individual moderation criteria. These findings call for the development of evaluation methods that explicitly measure criterion-conditioned behaviour.

Deep learning models frequently encounter feature uncertainty in diverse learning scenarios, significantly impacting their performance and reliability. This challenge is particularly complex in multi-modal scenarios, where models must integrate information from different sources with inherent uncertainties. We propose Dynamic Uncertainty-Aware Learning (DUAL), a unified framework that effectively handles feature uncertainty in both single-modal and multi-modal scenarios. DUAL introduces three key innovations: Dynamic Feature Uncertainty Modeling, which continuously refines uncertainty estimates through joint consideration of feature characteristics and learning dynamics; Adaptive Distribution-Aware Modulation, which maintains balanced feature distributions through dynamic sample influence adjustment; and Uncertainty-aware Cross-Modal Relationship Learning, which explicitly models uncertainties in cross-modal interactions. Through extensive experiments, we demonstrate DUAL's effectiveness across multiple domains: in computer vision tasks, it achieves substantial improvements of 7.1 CIFAR-100, and 2.3 demonstrates consistent gains of 4.1 on CMU-MOSI for sentiment analysis, while achieving 1.4 on MISR. The code will be available on GitHub soon.

As generative models become increasingly diverse and powerful, cross-generator detection has emerged as a new challenge. Existing detection methods often memorize artifacts of specific generative models rather than learning transferable cues, leading to substantial failures on unseen generators. Surprisingly, this work finds that frozen visual foundation models, especially DINOv3, already exhibit strong cross-generator detection capability without any fine-tuning. Through systematic studies on frequency, spatial, and token perspectives, we observe that DINOv3 tends to rely on global, low-frequency structures as weak but transferable authenticity cues instead of high-frequency, generator-specific artifacts. Motivated by this insight, we introduce a simple, training-free token-ranking strategy followed by a lightweight linear probe to select a small subset of authenticity-relevant tokens. This token subset consistently improves detection accuracy across all evaluated datasets. Our study provides empirical evidence and a feasible hypothesis for understanding why foundation models generalize across diverse generators, offering a universal, efficient, and interpretable baseline for image forgery detection.

Transformers have demonstrated exceptional in-context learning (ICL) capabilities, enabling applications across natural language processing, computer vision, and sequential decision-making. In reinforcement learning, ICL reframes learning as a supervised problem, facilitating task adaptation without parameter updates. Building on prior work leveraging transformers for sequential decision-making, we propose Heuristic Transformer (HT), an in-context reinforcement learning (ICRL) approach that augments the in-context dataset with a belief distribution over rewards to achieve better decision-making. Using a variational auto-encoder (VAE), a low-dimensional stochastic variable is learned to represent the posterior distribution over rewards, which is incorporated alongside an in-context dataset and query states as prompt to the transformer policy. We assess the performance of HT across the Darkroom, Miniworld, and MuJoCo environments, showing that it consistently surpasses comparable baselines in terms of both effectiveness and generalization. Our method presents a promising direction to bridge the gap between belief-based augmentations and transformer-based decision-making.

Recent advances in AI-powered generative models have enabled the creation of increasingly realistic synthetic images, posing significant risks to information integrity and public trust on social media platforms. While robust detection frameworks and diverse, large-scale datasets are essential to mitigate these risks, existing academic efforts remain limited in scope: current datasets lack the diversity, scale, and realism required for social media contexts, while detection methods struggle with generalization to unseen generative technologies. To bridge this gap, we introduce So-Fake-Set, a comprehensive social media-oriented dataset with over 2 million high-quality images, diverse generative sources, and photorealistic imagery synthesized using 35 state-of-the-art generative models. To rigorously evaluate cross-domain robustness, we establish a novel and large-scale (100K) out-of-domain benchmark (So-Fake-OOD) featuring synthetic imagery from commercial models explicitly excluded from the training distribution, creating a realistic testbed for evaluating real-world performance. Leveraging these resources, we present So-Fake-R1, an advanced vision-language framework that employs reinforcement learning for highly accurate forgery detection, precise localization, and explainable inference through interpretable visual rationales. Extensive experiments show that So-Fake-R1 outperforms the second-best method, with a 1.3

In commonsense generation, given a set of input concepts, a model must generate a response that is not only commonsense bearing, but also capturing multiple diverse viewpoints. Numerous evaluation metrics based on form- and content-level overlap have been proposed in prior work for evaluating the diversity of a commonsense generation model. However, it remains unclear as to which metrics are best suited for evaluating the diversity in commonsense generation. To address this gap, we conduct a systematic meta-evaluation of diversity metrics for commonsense generation. We find that form-based diversity metrics tend to consistently overestimate the diversity in sentence sets, where even randomly generated sentences are assigned overly high diversity scores. We then use an Large Language Model (LLM) to create a novel dataset annotated for the diversity of sentences generated for a commonsense generation task, and use it to conduct a meta-evaluation of the existing diversity evaluation metrics. Our experimental results show that content-based diversity evaluation metrics consistently outperform the form-based counterparts, showing high correlations with the LLM-based ratings. We recommend that future work on commonsense generation should use content-based metrics for evaluating the diversity of their outputs.

Geometry optimization of atomic structures is a common and crucial task in computational chemistry and materials design. Following the learning to optimize paradigm, we propose a new multi-agent reinforcement learning method called Multi-Agent Crystal Structure optimization (MACS) to address periodic crystal structure optimization. MACS treats geometry optimization as a partially observable Markov game in which atoms are agents that adjust their positions to collectively discover a stable configuration. We train MACS across various compositions of reported crystalline materials to obtain a policy that successfully optimizes structures from the training compositions as well as structures of larger sizes and unseen compositions, confirming its excellent scalability and zero-shot transferability. We benchmark our approach against a broad range of state-of-the-art optimization methods and demonstrate that MACS optimizes periodic crystal structures significantly faster, with fewer energy calculations, and the lowest failure rate.

The rapid advancement of generative models in creating highly realistic images poses substantial risks for misinformation dissemination. For instance, a synthetic image, when shared on social media, can mislead extensive audiences and erode trust in digital content, resulting in severe repercussions. Despite some progress, academia has not yet created a large and diversified deepfake detection dataset for social media, nor has it devised an effective solution to address this issue. In this paper, we introduce the Social media Image Detection dataSet (SID-Set), which offers three key advantages: (1) extensive volume, featuring 300K AI-generated/tampered and authentic images with comprehensive annotations, (2) broad diversity, encompassing fully synthetic and tampered images across various classes, and (3) elevated realism, with images that are predominantly indistinguishable from genuine ones through mere visual inspection. Furthermore, leveraging the exceptional capabilities of large multimodal models, we propose a new image deepfake detection, localization, and explanation framework, named SIDA (Social media Image Detection, localization, and explanation Assistant). SIDA not only discerns the authenticity of images, but also delineates tampered regions through mask prediction and provides textual explanations of the model's judgment criteria. Compared with state-of-the-art deepfake detection models on SID-Set and other benchmarks, extensive experiments demonstrate that SIDA achieves superior performance among diversified settings. The code, model, and dataset will be released.

The scarcity of labeled training samples poses a significant challenge in hyperspectral image classification. Cross-scene classification has been shown to be an effective approach to tackle the problem of limited sample learning. This paper investigates the usage of generative adversarial networks (GANs) to enable collaborative artificial intelligence learning on hyperspectral datasets. We propose and design a specialized architecture, termed Hyper-CycleGAN, for heterogeneous transfer learning across source and target scenes. This architecture enables the establishment of bidirectional mappings through efficient adversarial training and merges both source-to-target and target-to-source generators. The proposed Hyper-CycleGAN architecture harnesses the strengths of GANs, along with custom modifications like the integration of multi-scale attention mechanisms to enhance feature learning capabilities specifically tailored for hyperspectral data. To address training instability, the Wasserstein generative adversarial network with gradient penalty (WGAN-GP) loss discriminator is utilized. Additionally, a label smoothing technique is introduced to enhance the generalization capability of the generator, particularly in handling unlabeled samples, thus improving model robustness. Experimental results are performed to validate and confirm the effectiveness of the cross-domain Hyper-CycleGAN approach by demonstrating its applicability to two real-world cross-scene hyperspectral image datasets. Addressing the challenge of limited labeled samples in hyperspectral image classification, this research makes significant contributions and gives valuable insights for remote sensing, environmental monitoring, and medical imaging applications.

Transformer architectures have become popular across deep-learning disciplines due to their capability of efficiently integrating information across extensive temporal spans and handling large datasets. Recently, this property of transformer models has also been utilized for reinforcement learning (RL) by learning in-context. In in-context learning for decision-making problems, i.e., RL, a transformer model is usually pre-trained on an offline dataset and is tasked to predict the most likely action given a context. Such a model is able to make inference on the fly without parameter updates. Despite great success, the use of transformer architectures for RL is still in its infancy. In this paper, we further investigate the in-context learning abilities of transformer-based goal-oriented RL. We introduce Goal-Focused Transformer (GFT), a transformer meta-agent for goal-oriented RL. Building upon the Decision-Pretrained Transformer (DPT), GFT incorporates a function which distills goal information from the context, which we refer to as "goal-controller" (gc) and facilitates task inference during evaluation. By learning to distil useful information from context about the goal states, GFT enhances the exploration-exploitation dynamics and achieves superior performance and stability compared to DPT in environments with sparse rewards. Our contributions highlight GFT's efficacy in increasing average return, enhancing data efficiency, and providing a valuable mechanism for operating in dynamic environments while consistently striving to achieve predefined objectives.

This paper investigates how deep multi-agent reinforcement learning can enable the scalable and privacy-preserving coordination of residential energy flexibility. The coordination of distributed resources such as electric vehicles and heating will be critical to the successful integration of large shares of renewable energy in our electricity grid and, thus, to help mitigate climate change. The pre-learning of individual reinforcement learning policies can enable distributed control with no sharing of personal data required during execution. However, previous approaches for multi-agent reinforcement learning-based distributed energy resources coordination impose an ever greater training computational burden as the size of the system increases. We therefore adopt a deep multi-agent actor-critic method which uses a \emph{centralised but factored critic} to rehearse coordination ahead of execution. Results show that coordination is achieved at scale, with minimal information and communication infrastructure requirements, no interference with daily activities, and privacy protection. Significant savings are obtained for energy users, the distribution network and greenhouse gas emissions. Moreover, training times are nearly 40 times shorter than with a previous state-of-the-art reinforcement learning approach without the factored critic for 30 homes.

ChatGPT, as a language model based on large-scale pre-training, has exerted a profound influence on the domain of machine translation. In ChatGPT, a "Prompt" refers to a segment of text or instruction employed to steer the model towards generating a specific category of response. The design of the translation prompt emerges as a key aspect that can wield influence over factors such as the style, precision and accuracy of the translation to a certain extent. However, there is a lack of a common standard and methodology on how to design and select a translation prompt. Accordingly, this paper proposes a generic taxonomy, which defines gradable translation prompts in terms of expression type, translation style, POS information and explicit statement, thus facilitating the construction of prompts endowed with distinct attributes tailored for various translation tasks. Specific experiments and cases are selected to validate and illustrate the effectiveness of the method.

Generative Commonsense Reasoning (GCR) requires a model to reason about a situation using commonsense knowledge, while generating coherent sentences. Although the quality of the generated sentences is crucial, the diversity of the generation is equally important because it reflects the model's ability to use a range of commonsense knowledge facts. Large Language Models (LLMs) have shown proficiency in enhancing the generation quality across various tasks through in-context learning (ICL) using given examples without the need for any fine-tuning. However, the diversity aspect in LLM outputs has not been systematically studied before. To address this, we propose a simple method that diversifies the LLM generations, while preserving their quality. Experimental results on three benchmark GCR datasets show that our method achieves an ideal balance between the quality and diversity. Moreover, the sentences generated by our proposed method can be used as training data to improve diversity in existing commonsense generators.

To achieve the desired quality standards of certain manufactured materials, the involved parameters are still adjusted by knowledge-based procedures according to human expertise, which can be costly and time-consuming. To optimize operational efficiency and provide decision support for human experts, we develop a general continuous control framework that utilizes deep reinforcement learning (DRL) to automatically determine the main control parameters, in situations where simulation environments are unavailable and traditional PID controllers are not viable options. In our work, we aim to automatically learn the key control parameters to achieve the desired outlet thickness of the manufactured material. We first construct a surrogate environment based on real-world expert trajectories obtained from the true underlining manufacturing process to achieve this. Subsequently, we train a DRL agent within the surrogate environment. Our results suggest a Proximal Policy Optimization (PPO) algorithm combined with a Multi-Layer Perceptron (MLP) surrogate environment to successfully learn a policy that continuously changes parameter configurations optimally, achieving the desired target material thickness within an acceptable range.

Prior work has shown that the ordering in which concepts are shown to a commonsense generator plays an important role, affecting the quality of the generated sentence. However, it remains a challenge to determine the optimal ordering of a given set of concepts such that a natural sentence covering all the concepts could be generated from a pretrained generator. To understand the relationship between the ordering of the input concepts and the quality of the generated sentences, we conduct a systematic study considering multiple language models (LMs) and concept ordering strategies. We find that BART-large model consistently outperforms all other LMs considered in this study when fine-tuned using the ordering of concepts as they appear in CommonGen training data as measured using multiple evaluation metrics. Moreover, the larger GPT3-based large language models (LLMs) variants do not necessarily outperform much smaller LMs on this task, even when fine-tuned on task-specific training data. Interestingly, human annotators significantly reorder input concept sets when manually writing sentences covering those concepts, and this ordering provides the best sentence generations independently of the LM used for the generation, outperforming a probabilistic concept ordering baseline

Relative overgeneralization (RO) is a pathology that can arise in cooperative multi-agent tasks when the optimal joint action's utility falls below that of a sub-optimal joint action. RO can cause the agents to get stuck into local optima or fail to solve cooperative tasks requiring significant coordination between agents within a given timestep. In this work, we empirically find that, in multi-agent reinforcement learning (MARL), both value-based and policy gradient MARL algorithms can suffer from RO and fail to learn effective coordination policies. To better overcome RO, we propose a novel approach called curriculum learning for relative overgeneralization (CURO). To solve a target task that exhibits strong RO, in CURO, we first fine-tune the reward function of the target task to generate source tasks to train the agent. Then, to effectively transfer the knowledge acquired in one task to the next, we use a transfer learning method that combines value function transfer with buffer transfer, which enables more efficient exploration in the target task. CURO is general and can be applied to both value-based and policy gradient MARL methods. We demonstrate that, when applied to QMIX, HAPPO, and HATRPO, CURO can successfully overcome severe RO, achieve improved performance, and outperform baseline methods in a variety of challenging cooperative multi-agent tasks.