Standard policy gradients weight each sampled action by advantage alone, regardless of how likely that action was under the current policy. This creates two pathologies: within a single decision context (e.g. one image or prompt), a rare negative-advantage action can disproportionately distort the update direction; across many such contexts in a batch, the expected gradient over-allocates budget to contexts the policy already handles well. We introduce the \textit{Delightful Policy Gradient} (DG), which gates each term with a sigmoid of \emph{delight}, the product of advantage and action surprisal (negative log-probability). For $K$-armed bandits, DG provably improves directional accuracy in a single context and, across multiple contexts, shifts the expected gradient strictly closer to the supervised cross-entropy oracle. This second effect is not variance reduction: it persists even with infinite samples. Empirically, DG outperforms REINFORCE, PPO, and advantage-weighted baselines across MNIST, transformer sequence modeling, and continuous control, with larger gains on harder tasks.

Kernel methods are widely used in causal inference for tasks such as treatment effect estimation, policy evaluation, and policy learning. The bootstrap is a standard tool for uncertainty quantification because of its broad applicability. As increasingly large datasets become available, such as the 2023 U.S. Natality data from the National Vital Statistics System (NVSS), which includes 3,596,017 registered births, the computational demands of these methods increase substantially. Kernel methods are known to scale poorly with sample size, and this limitation is further exacerbated by the repeated re-fitting required by the bootstrap. As a result, bootstrap-based inference for kernel-based estimators can become computationally infeasible in large-scale settings. In this paper, we address these challenges by extending the causal Bag of Little Bootstraps (cBLB) algorithm to kernel methods. Our approach achieves computational scalability by combining subsampling and resampling while preserving first-order uncertainty quantification and asymptotically correct coverage. We evaluate the method across three representative implementations: kernelized augmented outcome-weighted learning, kernel-based minimax weighting, and double machine learning with kernel support vector machines. We show in simulations that our method yields confidence intervals with nominal coverage at a fraction of the computational cost. We further demonstrate its utility in a real-world application by estimating the effect of any amount of smoking on birth weight, as well as the optimal treatment regime, using the NVSS dataset, where the standard bootstrap is prohibitively expensive computationally and effectively infeasible at this scale.

Simulation plays a central role in scientific discovery. In many applications, the bottleneck is no longer running a simulator; it is choosing among large families of plausible simulators, each corresponding to different forward models/hypotheses consistent with observations. Over large model families, classical Bayesian workflows for model selection are impractical. Furthermore, amortized model selection methods typically hard-code a fixed model prior or complexity penalty at training time, requiring users to commit to a particular parsimony assumption before seeing the data. We introduce PRISM, a simulation-based encoder-decoder that infers a joint posterior over both discrete model structures and associated continuous parameters, while enabling test-time control of model complexity via a tunable model prior that the network is conditioned on. We show that PRISM scales to families with combinatorially many (up to billions) of model instantiations on a synthetic symbolic regression task. As a scientific application, we evaluate PRISM on biophysical modeling for diffusion MRI data, showing the ability to perform model selection across several multi-compartment models, on both synthetic and in vivo neuroimaging data.

Imbalanced datasets pose a difficulty in fraud detection, as classifiers are often biased toward the majority class and perform poorly on rare fraudulent transactions. Synthetic data generation is therefore commonly used to mitigate this problem. In this work, we propose the Clustered Embedding Diffusion-Transformer (EmDT), a diffusion model designed to generate fraudulent samples. Our key innovation is to leverage UMAP clustering to identify distinct fraudulent patterns, and train a Transformer denoising network with sinusoidal positional embeddings to capture feature relationships throughout the diffusion process. Once the synthetic data has been generated, we employ a standard decision-tree-based classifier (e.g., XGBoost) for classification, as this type of model remains better suited to tabular datasets. Experiments on a credit card fraud detection dataset demonstrate that EmDT significantly improves downstream classification performance compared to existing oversampling and generative methods, while maintaining comparable privacy protection and preserving feature correlations present in the original data.

Morphology-control co-design concerns the coupled optimization of an agent's body structure and control policy. This problem exhibits a bi-level structure, where the control dynamically adapts to the morphology to maximize performance. Existing methods typically neglect the control's adaptation dynamics by adopting a single-level formulation that treats the control policy as fixed when optimizing morphology. This can lead to inefficient optimization, as morphology updates may be misaligned with control adaptation. In this paper, we revisit the co-design problem from a game-theoretic perspective, modeling the intrinsic coupling between morphology and control as a novel variant of a Stackelberg game. We propose Stackelberg Proximal Policy Optimization (Stackelberg PPO), which explicitly incorporates the control's adaptation dynamics into morphology optimization. By modeling this intrinsic coupling, our method aligns morphology updates with control adaptation, thereby stabilizing training and improving learning efficiency. Experiments across diverse co-design tasks demonstrate that Stackelberg PPO outperforms standard PPO in both stability and final performance, opening the way for dramatically more efficient robotics designs.

We study the following question: given an efficient approximation algorithm for an optimization problem, can we learn efficiently in the same setting? We give a formal affirmative answer to this question in the form of a reduction from online learning to offline approximate optimization using an efficient algorithm that guarantees near optimal regret. The algorithm is efficient in terms of the number of oracle calls to a given approximation oracle - it makes only logarithmically many such calls per iteration.

We design differentially private learning algorithms that are agnostic to the learning model assuming access to limited amount of unlabeled public data. First, we give a new differentially private algorithm for answering a sequence of $m$ online classification queries (given by a sequence of $m$ unlabeled public feature vectors) based on a private training set. Our private algorithm follows the paradigm of subsample-and-aggregate, in which any generic non-private learner is trained on disjoint subsets of the private training set, then for each classification query, the votes of the resulting classifiers ensemble are aggregated in a differentially private fashion.

The accurate exposure is the key of capturing high-quality photos in computational photography, especially for mobile phones that are limited by sizes of camera modules. Inspired by luminosity masks usually applied by professional photographers, in this paper, we develop a novel algorithm for learning local exposures with deep reinforcement adversarial learning. To be specific, we segment an image into sub-images that can reflect variations of dynamic range exposures according to raw low-level features. Based on these sub-images, a local exposure for each sub-image is automatically learned by virtue of policy network sequentially while the reward of learning is globally designed for striking a balance of overall exposures. The aesthetic evaluation function is approximated by discriminator in generative adversarial networks. The reinforcement learning and the adversarial learning are trained collaboratively by asynchronous deterministic policy gradient and generative loss approximation. To further simply the algorithmic architecture, we also prove the feasibility of leveraging the discriminator as the value function. Further more, we employ each local exposure to retouch the raw input image respectively, thus delivering multiple retouched images under different exposures which are fused with exposure blending. The extensive experiments verify that our algorithms are superior to state-of-the-art methods in terms of quantitative accuracy and visual illustration.

Australia's eSafety commissioner wrote to X in January after its AI chatbot Grok was used to generate sexualised images of women and children online. Australia's eSafety commissioner wrote to X in January after its AI chatbot Grok was used to generate sexualised images of women and children online. Child abuse material'systemic' on Elon Musk's X amid Grok scandal, Australian online safety regulator warned The Australian online safety regulator warned Elon Musk's X amid the Grok sexualised image generation scandal that it found child abuse material was "particularly systemic" on X and more accessible than on "any other mainstream service", correspondence obtained by Guardian Australia reveals. The eSafety commissioner wrote to X in January after its chatbot, Grok, was used to generate sexualised images of women and children online, which the prime minister, Anthony Albanese, described as "abhorrent". In the letter, obtained by Guardian Australia under freedom of information laws, eSafety's general manager of regulatory operations, Heidi Snell, pointed to Musk's promise when taking over the platform in 2022 that "removing child exploitation is priority #1", but said "the availability of CSEM [child sexual exploitation material] continues to appear particularly systemic on X".

We present an end-to-end framework for solving the Vehicle Routing Problem (VRP) using reinforcement learning. In this approach, we train a single policy model that finds near-optimal solutions for a broad range of problem instances of similar size, only by observing the reward signals and following feasibility rules. We consider a parameterized stochastic policy, and by applying a policy gradient algorithm to optimize its parameters, the trained model produces the solution as a sequence of consecutive actions in real time, without the need to re-train for every new problem instance. On capacitated VRP, our approach outperforms classical heuristics and Google's OR-Tools on medium-sized instances in solution quality with comparable computation time (after training). We demonstrate how our approach can handle problems with split delivery and explore the effect of such deliveries on the solution quality.