Beneath this caution lingers a fundamental question about whether these annual gatherings are still worth it, given Mr Trump's clear disdain. He prefers bilateral dealmaking to multilateral consensus-building. This is the president's first such foray onto the world stage since his inauguration and his six partners will be looking anxiously to see whether he wants to pick a fight - or look statesmanlike - for voters back home. Max Bergmann, director of the Europe, Russia and Eurasia Program at the Center for Strategic and International Studies, said: "The question now is not so much'is this an awkward family gathering?' I think the question is: 'is this still a family?'"

The Mossad published footage of its operatives carrying out covert actions inside Iran prior to Israel's preemptive attack. Israel's overnight strike on Iran was not only one of the most ambitious aerial campaigns in recent history, it was the result of years of covert planning, surveillance and infiltration by Israeli intelligence. While dozens of fighter jets bombed nuclear and military targets across Iran early Friday morning, the groundwork had long been laid by Mossad agents working in lockstep with the Israeli military. Code-named "Am Kelavi" (Rising Lion), the preemptive operation was the product of unprecedented coordination between the Israeli air force, the Military Intelligence Directorate, Mossad and the country's defense industries. For years, they worked "shoulder to shoulder" to gather the intelligence files needed to eliminate Iran's most sensitive military and nuclear assets.

This paper is a brief introduction to our recently initiated project named VERIFAI: Traceability and verification of natural language requirements. The project addresses the challenges in the traceability and verification of formal specifications through providing support for the automatic generation of the formal specifications and the traceability of the requirements from the initial software design stage through the systems implementation and verification. Approaches explored in this project include Natural Language Processing, use of ontologies to describe the software system domain, reuse of existing software artefacts from similar systems (i.e. through similarity based reuse) and large language models to identify and declare the specifications as well as use of artificial intelligence to guide the process.

Assistance games are a promising alternative to reinforcement learning from human feedback (RLHF) for training AI assistants. Assistance games resolve key drawbacks of RLHF, such as incentives for deceptive behavior, by explicitly modeling the interaction between assistant and user as a two-player game where the assistant cannot observe their shared goal. Despite their potential, assistance games have only been explored in simple settings. Scaling them to more complex environments is difficult because it requires both solving intractable decision-making problems under uncertainty and accurately modeling human users' behavior. We present the first scalable approach to solving assistance games and apply it to a new, challenging Minecraft-based assistance game with over $10^{400}$ possible goals. Our approach, AssistanceZero, extends AlphaZero with a neural network that predicts human actions and rewards, enabling it to plan under uncertainty. We show that AssistanceZero outperforms model-free RL algorithms and imitation learning in the Minecraft-based assistance game. In a human study, our AssistanceZero-trained assistant significantly reduces the number of actions participants take to complete building tasks in Minecraft. Our results suggest that assistance games are a tractable framework for training effective AI assistants in complex environments. Our code and models are available at https://github.com/cassidylaidlaw/minecraft-building-assistance-game.

Imagine trying to find a hidden k -vertex clique (fully connected subgraph) within an otherwise random n -vertex graph (network). While it is possible to find a hidden clique of size k log n by brute-force search, all known "fast" (polynomial-time) algorithms only work if the clique is much larger: k n . Is this an inherent limitation of fast algorithms or should we continue looking for a better one? Similar questions of computational complexity arise in many other statistical settings, such as community detection, clustering, and sparse PCA. While we lack the tools to prove definitively that fast algorithms require k n, this survey describes one sense in which we can prove this threshold is fundamental: all algorithms based on low-degree polynomials -- for instance, counting triangles in the graph would be a degree-3 polynomial -- provably fail (in an appropriate sense) when k n . Furthermore, these low-degree algorithms tend to capture the best tools in our algorithmic toolkit for problems of this style, so finding a fast algorithm for k n would seem to require a major breakthrough or may simply be impossible. This provides a lens for predicting and explaining the limitations of fast algorithms across many different settings.

Diffusion bridges are a promising class of deep-learning methods for sampling from unnormalized distributions. Recent works show that the Log Variance (LV) loss consistently outperforms the reverse Kullback-Leibler (rKL) loss when using the reparametrization trick to compute rKL-gradients. While the on-policy LV loss yields identical gradients to the rKL loss when combined with the log-derivative trick for diffusion samplers with non-learnable forward processes, this equivalence does not hold for diffusion bridges or when diffusion coefficients are learned. Based on this insight we argue that for diffusion bridges the LV loss does not represent an optimization objective that can be motivated like the rKL loss via the data processing inequality. Our analysis shows that employing the rKL loss with the log-derivative trick (rKL-LD) does not only avoid these conceptual problems but also consistently outperforms the LV loss. Experimental results with different types of diffusion bridges on challenging benchmarks show that samplers trained with the rKL-LD loss achieve better performance. From a practical perspective we find that rKL-LD requires significantly less hyperparameter optimization and yields more stable training behavior.

Reinforcement Learning Finetuning (RFT) has significantly advanced the reasoning capabilities of large language models (LLMs) by enabling long chains of thought, self-correction, and effective tool use. While recent works attempt to extend RFT to vision-language models (VLMs), these efforts largely produce text-only reasoning conditioned on static image inputs, falling short of true multimodal reasoning in the response. In contrast, test-time methods like Visual Sketchpad incorporate visual steps but lack training mechanisms. We introduce VTool-R1, the first framework that trains VLMs to generate multimodal chains of thought by interleaving text and intermediate visual reasoning steps. VTool-R1 integrates Python-based visual editing tools into the RFT process, enabling VLMs to learn when and how to generate visual reasoning steps that benefit final reasoning. Trained with outcome-based rewards tied to task accuracy, our approach elicits strategic visual tool use for reasoning without relying on process-based supervision. Experiments on structured visual question answering over charts and tables show that VTool-R1 enhances reasoning performance by teaching VLMs to "think with images" and generate multimodal chain of thoughts with tools.

In this paper we present a sample treebank for Old English based on the UD Cairo sentences, collected and annotated as part of a classroom curriculum in Historical Linguistics. To collect the data, a sample of 20 sentences illustrating a range of syntactic constructions in the world's languages, we employ a combination of LLM prompting and searches in authentic Old English data. For annotation we assigned sentences to multiple students with limited prior exposure to UD, whose annotations we compare and adjudicate. Our results suggest that while current LLM outputs in Old English do not reflect authentic syntax, this can be mitigated by post-editing, and that although beginner annotators do not possess enough background to complete the task perfectly, taken together they can produce good results and learn from the experience. We also conduct preliminary parsing experiments using Modern English training data, and find that although performance on Old English is poor, parsing on annotated features (lemma, hyperlemma, gloss) leads to improved performance.

Large Language models (LLMs) are transforming digital usage, particularly in text generation, image creation, information retrieval and code development. ChatGPT, launched by OpenAI in November 2022, quickly became a reference, prompting the emergence of competitors such as Google's Gemini. However, these technological advances raise new cybersecurity challenges, including prompt injection attacks, the circumvention of regulatory measures ( jailbreaking), the spread of misinformation (hallucinations) and risks associated with deep fakes. This paper presents a comparative analysis of the security and alignment levels of ChatGPT and Gemini, as well as a taxonomy of jailbreak techniques associated with experiments.

Language Models (LMs) are prone to memorizing parts of their data during training and unintentionally emitting them at generation time, raising concerns about privacy leakage and disclosure of intellectual property. While previous research has identified properties such as context length, parameter size, and duplication frequency, as key drivers of unintended memorization, little is known about how the latent structure modulates this rate of memorization. We investigate the role of Intrinsic Dimension (ID), a geometric proxy for the structural complexity of a sequence in latent space, in modulating memorization. Our findings suggest that ID acts as a suppressive signal for memorization: compared to low-ID sequences, high-ID sequences are less likely to be memorized, particularly in overparameterized models and under sparse exposure. These findings highlight the interaction between scale, exposure, and complexity in shaping memorization.