We consider the privacy amplification properties of a sampling scheme in which a user's data is used in k steps chosen randomly and uniformly from a sequence (or set) of t steps. This sampling scheme has been recently applied in the context of differentially private optimization [Chua et al., 2024a, Choquette-Choo et al., 2025] and is also motivated by communication-efficient high-dimensional private aggregation [Asi et al., 2025]. Existing analyses of this scheme either rely on privacy amplification by shuffling which leads to overly conservative bounds or require Monte Carlo simulations that are computationally prohibitive in most practical scenarios. We give the first theoretical guarantees and numerical estimation algorithms for this sampling scheme. In particular, we demonstrate that the privacy guarantees of random k-out-of-t allocation can be upper bounded by the privacy guarantees of the well-studied independent (or Poisson) subsampling in which each step uses the user's data with probability (1+o(1))k/t. Further, we provide two additional analysis techniques that lead to numerical improvements in several parameter regimes. Altogether, our bounds give efficiently-computable and nearly tight numerical results for random allocation applied to Gaussian noise addition.

Graph coarsening aims to diminish the size of a graph to lighten its memory footprint, and has numerous applications in graph signal processing and machine learning. It is usually defined using a reduction matrix and a lifting matrix, which, respectively, allows to project a graph signal from the original graph to the coarsened one and back. This results in a loss of information measured by the so-called Restricted Spectral Approximation (RSA). Most coarsening frameworks impose a fixed relationship between the reduction and lifting matrices, generally as pseudoinverses of each other, and seek to define a coarsening that minimizes the RSA. In this paper, we remark that the roles of these two matrices are not entirely symmetric: indeed, putting constraints on the lifting matrix alone ensures the existence of important objects such as the coarsened graph's adjacency matrix or Laplacian.

Transformers have theoretical limitations in modeling certain sequence-to-sequence tasks, yet it remains largely unclear if these limitations play a role in large-scale pretrained LLMs, or whether LLMs might effectively overcome these constraints in practice due to the scale of both the models themselves and their pretraining data. We explore how these architectural constraints manifest after pretraining, by studying a family of retrieval and copying tasks inspired by Liu et al. [2024a]. We use a recently proposed framework for studying length generalization [Huang et al., 2025] to provide guarantees for each of our settings.

While foundation models have shown promise across a variety of fields, astronomy lacks a unified framework for joint modeling across its highly diverse data modalities.

GTA 6 - all you need to know about Rockstar's blockbuster game The latest instalment in Rockstar's blockbuster game franchise, Grand Theft Auto, is set to be the biggest games launch of the year. Details are still scant, although we do now know that GTA 6 will be available to pre-order on 25 June, the developer has announced . Analysts believe Rockstar's action adventure could become the most expensive game ever made, with estimates putting development costs at more than $1bn (£866m). We're still awaiting some crucial information about the game - but here's what we do and don't know about GTA 6 so far. When is GTA 6 coming out?

Inspired by the success of DeepSeek-R1, we explore the potential of rule-based reinforcement learning (RL) in MLLM post-training for perception policy learning. While promising, our initial experiments reveal that incorporating a thinking process through RL does not consistently lead to performance gains across all visual perception tasks. This leads us to delve into the essential role of RL in the context of visual perception. In this work, we return to the fundamentals and explore the effects of RL on different perception tasks. We observe that the perceptual perplexity is a major factor in determining the effectiveness of RL. We also observe that reward design plays a crucial role in further approaching the upper limit of model perception. To leverage these findings, we propose Perception-R1, a scalable RL framework using GRPO during MLLM post-training.

Try One of macOS 27's Best Features Right Now Apple's fall macOS release will let you build Shortcuts by typing what you want to happen. But Claude Code and Codex users don't have to wait. Buried deep inside everything announced at WWDC this year was something I, an Apple Shortcuts enthusiast, can't wait to try: the ability to make Apple Shortcuts using generative artificial intelligence. In macOS 27, you'll be able to just type what you want a shortcut to do, and the app will build it. Anyone who builds shortcuts regularly knows the process of doing so can be tedious, even if the end results save you a lot of time.

The biggest World Cup ever is pushing fans, players, and host cities to their limits--and experts say this is only the beginning. It's often said that bigger means better. This year's FIFA World Cup may put that to the test. By almost any metric, the 2026 tournament is the largest ever: the most host countries; the longest distances between stadiums; the most players, teams, and matches; and then there's the eye-watering ticket prices . The scale is a logistical nightmare for fans, teams, and host cities. Held across three countries-- Canada, Mexico, and the US--48 teams (up from the usual 32) will navigate 16 host cities separated by thousands of miles and four distinct time zones.

Side-channel analysis (SCA) poses a real-world threat by exploiting unintentional physical signals to extract secret information from secure devices. Evaluation labs also use the same techniques to certify device security. In recent years, deep learning has emerged as a prominent method for SCA, achieving state-ofthe-art attack performance at the cost of interpretability. Understanding how neural networks extract secrets is crucial for security evaluators aiming to defend against such attacks, as only by understanding the attack can one propose better countermeasures. In this work, we apply mechanistic interpretability to neural networks trained for SCA, revealing how models exploit what leakage in side-channel traces. We focus on sudden jumps in performance to reverse engineer learned representations, ultimately recovering secret masks and moving the evaluation process from blackbox to white-box. Our results show that mechanistic interpretability can scale to realistic SCA settings, even when relevant inputs are sparse, model accuracies are low, and side-channel protections prevent standard input interventions.

UK's top data and AI regulator quits after'inappropriate' humour John Edwards, the UK's information commissioner, has resigned following a workplace investigation. I have accepted that there have been occasions where I exercised poor judgement and made attempts at humour that were inappropriate and caused offence, he said in a statement on Friday. The Information Commissioner's Office (ICO) is responsible for regulating AI in the UK and also oversees data protection regulation and the freedom of information law. Edwards' resignation was confirmed by the government, which said it had come after an independent probe that took place regarding allegations made against him. The government expects the highest standards of conduct from all senior leaders in public life, said a spokesperson for the Department for Science, Innovation and Technology (DSIT).