Stocks from the United States to Japan have risen sharply amid hopes that an end to the longest US government shutdown in history is imminent. US lawmakers on Sunday moved to end a five-week impasse over government funding, a boost for investors unnerved by signs of growing weakness in the US economy and the sky-high evaluations of firms involved in artificial intelligence. The funding package still needs to win final approval in the Senate and then pass the US House of Representatives, after which it would go to US President Donald Trump for his signature - a process expected to take days. Stock markets in the Asia Pacific made large gains on Monday, while futures in the US also rose in advance of stock exchanges reopening. South Korea's benchmark KOSPI led the gains, rising about 3 percent as of 4pm local time (07:00 GMT).

Doctors from Scotland and the US have completed what is thought to be a world-first stroke procedure using a robot. Prof Iris Grunwald, of the University of Dundee, performed the remote thrombectomy - the removal of blood clots after a stroke - on a human cadaver that had been donated to medical science. The professor was at Ninewells Hospital in Dundee, while the body she was operating on while using the machine was across the city at the university. Hours later, Ricardo Hanel - a neurosurgeon in Florida - used the technology to carry out the first transatlantic surgery from his Jacksonville base on a human body in Dundee over 4,000 miles (6,400km) away. The team has called it a potential game changer if it becomes approved for use on patients.

A woman who was plunged into sudden menopause after surgery to remove both ovaries is spearheading efforts to change NHS policy. Kate Dyson, 44, from Hastings, East Sussex, underwent the surgery six months ago after having a subtotal hysterectomy in 2021 to remove her uterus - a procedure which leaves the cervix in place. The mum-of-three says she was completely unprepared for the impact of surgical menopause, which is triggered by both ovaries being removed. Honestly, it was like falling off a hormonal cliff edge, she told BBC Radio Sussex. Within hours of the surgery I was home the same day.

Apple has removed two of the most popular gay dating apps in China from the App Store after receiving an order from China's main internet regulator and censorship authority, WIRED has learned. The move comes as reports of Blued and Finka disappearing from the iOS App Store and several Android app stores circulated on Chinese social media over the weekend. The apps appear to still be functional for users in the country who already have them downloaded. "We follow the laws in the countries where we operate. Based on an order from the Cyberspace Administration of China, we have removed these two apps from the China storefront only," an Apple spokesperson said in an email.

'Neuralink is doing legitimate technology development for neuroscience, and then Elon Musk comes along and starts talking about telepathy and stuff.' 'Neuralink is doing legitimate technology development for neuroscience, and then Elon Musk comes along and starts talking about telepathy and stuff.' Investors' 'dumb transhumanist ideas' setting back neurotech progress, say experts I t has been an excellent year for neurotech, if you ignore the people funding it. In August, a tiny brain implant successfully decoded the inner speech of paralysis patients. In October, an eye restored sight to patients who had lost their vision. It would just be better, say experts, if the most famous investors in the space - tech magnates such as Elon Musk and OpenAI's Sam Altman - were less interested in uploading their brains to computers or merging with AI. "It's distorting the debate a lot," said Marcello Ienca, a professor of neuroethics at the Technical University of Munich.

You probably don't give much thought to the device that you're reading this article on, as long as it looks good and keeps working. But the elements that power and run it are the subject of an escalating struggle between the world's two biggest economies - the US and China - with African countries in the eye of the storm. The African continent is rich in critical minerals and metals - like lithium, rare earths, cobalt and tungsten - which are vital to making and running our personal tech. Such materials are also essential for everything from electric vehicles, to AI data centres, and weapon systems. China has long been the biggest player in the global market for critical minerals and metals.

The personalization of black-box large language models (LLMs) is a critical yet challenging task. Existing approaches predominantly rely on context injection, where user history is embedded into the prompt to directly guide the generation process. However, this single-step paradigm imposes a dual burden on the model: generating accurate content while simultaneously aligning with user-specific styles. This often results in a trade-off that compromises output quality and limits precise control. To address this fundamental tension, we propose Reflective Personalization Optimization (RPO), a novel framework that redefines the personalization paradigm by decoupling content generation from alignment. RPO operates in two distinct stages: first, a base model generates a high-quality, generic response; then, an external reflection module explicitly rewrites this output to align with the user's preferences. This reflection module is trained using a two-stage process. Initially, supervised fine-tuning is employed on structured rewriting trajectories to establish a core personalized reasoning policy that models the transformation from generic to user-aligned responses. Subsequently, reinforcement learning is applied to further refine and enhance the quality of the personalized outputs. Comprehensive experiments on the LaMP benchmark demonstrate that RPO, by decoupling content generation from personalization, significantly outperforms state-of-the-art baselines. These findings underscore the superiority of explicit response shaping over implicit context injection. Moreover, RPO introduces an efficient, model-agnostic personalization layer that can be seamlessly integrated with any underlying base model, paving the way for a new and effective direction in user-centric generation scenarios.

Preferential Bayesian optimization allows optimization of objectives that are either expensive or difficult to measure directly, by relying on a minimal number of comparative evaluations done by a human expert. Generating candidate solutions for evaluation is also often expensive, but this cost is ignored by existing methods. We generalize preference-based optimization to explicitly account for production and evaluation costs with Consecutive Preferential Bayesian Optimization, reducing production cost by constraining comparisons to involve previously generated candidates. We also account for the perceptual ambiguity of the oracle providing the feedback by incorporating a Just-Noticeable Difference threshold into a probabilistic preference model to capture indifference to small utility differences. We adapt an information-theoretic acquisition strategy to this setting, selecting new configurations that are most informative about the unknown optimum under a preference model accounting for the perceptual ambiguity. We empirically demonstrate a notable increase in accuracy in setups with high production costs or with indifference feedback.

As predictive algorithms grow in popularity, using the same dataset to both train and test a new model has become routine across research, policy, and industry. Sample-splitting attains valid inference on model properties by using separate subsamples to estimate the model and to evaluate it. However, this approach has two drawbacks, since each task uses only part of the data, and different splits can lead to widely different estimates. Averaging across multiple splits, I develop an inference approach that uses more data for training, uses the entire sample for testing, and improves reproducibility. I address the statistical dependence from reusing observations across splits by proving a new central limit theorem for a large class of split-sample estimators under arguably mild and general conditions. Importantly, I make no restrictions on model complexity or convergence rates. I show that confidence intervals based on the normal approximation are valid for many applications, but may undercover in important cases of interest, such as comparing the performance between two models. I develop a new inference approach for such cases, explicitly accounting for the dependence across splits. Moreover, I provide a measure of reproducibility for p-values obtained from split-sample estimators. Finally, I apply my results to two important problems in development and public economics: predicting poverty and learning heterogeneous treatment effects in randomized experiments. I show that my inference approach with repeated cross-fitting achieves better power than previous alternatives, often enough to find statistical significance that would otherwise be missed.

Off-dynamics reinforcement learning (RL), where training and deployment transition dynamics are different, can be formulated as learning in a robust Markov decision process (RMDP) where uncertainties in transition dynamics are imposed. Existing literature mostly assumes access to generative models allowing arbitrary state-action queries or pre-collected datasets with a good state coverage of the deployment environment, bypassing the challenge of exploration. In this work, we study a more realistic and challenging setting where the agent is limited to online interaction with the training environment. To capture the intrinsic difficulty of exploration in online RMDPs, we introduce the supremal visitation ratio, a novel quantity that measures the mismatch between the training dynamics and the deployment dynamics. We show that if this ratio is unbounded, online learning becomes exponentially hard. We propose the first computationally efficient algorithm that achieves sublinear regret in online RMDPs with $f$-divergence based transition uncertainties. We also establish matching regret lower bounds, demonstrating that our algorithm achieves optimal dependence on both the supremal visitation ratio and the number of interaction episodes. Finally, we validate our theoretical results through comprehensive numerical experiments.