A new artificial intelligence tool designed to crack down on fraud has helped the UK government recover almost £500m over the last year, the BBC can reveal. More than a third of the money clawed back related to fraudulent activity during the Covid-19 pandemic, with other cash being recouped from unlawful council tax claims and illegal subletting of social housing. The government will announce later that a new AI tool which has helped to identify the fraud will now be licensed to other countries, including the US and Australia. Civil liberties campaigners have previously criticised the Labour government for its use of AI in trying to counter fraud. The Cabinet Office says the £480m recovered in the 12 months from April 2024 is the largest sum ever reclaimed by government anti-fraud teams in a single year.

Is recognising Palestine a way to'save face' for Western leaders? Organisers of the Global Sumud Flotilla, a Gaza-bound flotilla with pro-Palestinian activists on board carrying aid, reported hearing explosions and seeing multiple drone attacks from their boats situated off Greece from late Tuesday to the early hours of Wednesday. "Multiple drones, unidentified objects dropped, communications jammed and explosions heard from a number of boats," the Global Sumud Flotilla said in a statement, without adding whether there were any casualties. "We are witnessing these psychological operations firsthand, right now, but we will not be intimidated." Suited in a life jacket, Brazilian organiser Tiago Avila updated on his Instagram at midnight on Wednesday that a total of 10 attacks targeted multiple boats with sound bombs and explosive flares.

A South Korean court has handed a 15-year prison sentence to the boss of a lithium battery maker after a deadly fire last year. In June 2024, a blaze at a plant in Hwaseong city, about 45km (28 miles) south of the capital Seoul, killed 23 people, including 18 foreign workers, and injured eight others. The court found the blaze was an anticipated disaster and that Aricell chief executive Park Soon-kwan and other executives had caused the deaths of the workers. It is the longest jail term imposed under the country's industrial safety law, which punishes owners or bosses of firms with at least a year in prison, or fines of up to 1 billion won ($717,000; £530,000), for fatal incidents. Prosecutors had sought a 20-year term, arguing that company executives had made changes to the plant that meant it was difficult for workers to escape the fire.

An escalator used by Donald Trump abruptly stopped because of a safety mechanism that may have been triggered by his videographer, the United Nations has said. The videographer had been travelling backwards up the escalator to capture the US president's arrival with First Lady Melania Trump and may have inadvertently triggered the safety function upon reaching the top, a UN spokesperson said. Trump jokingly referred to the incident during his Tuesday speech at the UN building, saying: If the First Lady wasn't in great shape, she would've fallen. The White House had raised concerns that someone deliberately stopped the escalator as the couple were stepping on. If someone at the U.N. intentionally stopped the escalator as the President and First Lady were stepping on, they need to be fired and investigated immediately, White House press secretary Karoline Leavitt posted on X after the incident.

Tensor completion is an extension of matrix completion aimed at recovering a multiway data tensor by leveraging a given subset of its entries (observations) and the pattern of observation. The low-rank assumption is key in establishing a relationship between the observed and unobserved entries of the tensor. The low-rank tensor completion problem is typically solved using numerical optimization techniques, where the rank information is used either implicitly (in the rank minimization approach) or explicitly (in the error minimization approach). Current theories concerning these techniques often study probabilistic recovery guarantees under conditions such as random uniform observations and incoherence requirements. However, if an observation pattern exhibits some low-rank structure that can be exploited, more efficient algorithms with deterministic recovery guarantees can be designed by leveraging this structure. This work shows how to use only standard linear algebra operations to compute the tensor train decomposition of a specific type of ``fiber-wise" observed tensor, where some of the fibers of a tensor (along a single specific mode) are either fully observed or entirely missing, unlike the usual entry-wise observations. From an application viewpoint, this setting is relevant when it is easier to sample or collect a multiway data tensor along a specific mode (e.g., temporal). The proposed completion method is fast and is guaranteed to work under reasonable deterministic conditions on the observation pattern. Through numerical experiments, we showcase interesting applications and use cases that illustrate the effectiveness of the proposed approach.

Large language models (LLMs) often reflect Western-centric biases, limiting their effectiveness in diverse cultural contexts. Although some work has explored cultural alignment, the potential for cross-cultural transfer, using alignment in one culture to improve performance in others, remains underexplored. This paper investigates cross-cultural transfer of commonsense reasoning in the Arab world, where linguistic and historical similarities coexist with local cultural differences. Using a culturally grounded commonsense reasoning dataset covering 13 Arab countries, we evaluate lightweight alignment methods such as in-context learning and demonstration-based reinforcement (DITTO), alongside baselines like supervised fine-tuning and direct preference optimization. Our results show that merely 12 culture-specific examples from one country can improve performance in others by 10\% on average, within multilingual models. In addition, we demonstrate that out-of-culture demonstrations from Indonesia and US contexts can match or surpass in-culture alignment for MCQ reasoning, highlighting cultural commonsense transferability beyond the Arab world. These findings demonstrate that efficient cross-cultural alignment is possible and offer a promising approach to adapt LLMs to low-resource cultural settings.

Proper initialization is crucial for any system, particularly in multi-agent systems (MAS), where it plays a pivotal role in determining both the system's efficiency and effectiveness. However, existing MAS initialization methods do not fully account for the collaborative needs of the generated agents in subsequent stages. Inspired by the principles of effective team composition, we propose AgentInit, which aims to optimize the structure of agent teams. Specifically, in addition to multi-round interactions and reflections between agents during agent generation, AgentInit incorporates a Natural Language to Format mechanism to ensure consistency and standardization. Balanced team selection strategies using Pareto principles are subsequently applied to jointly consider agent team diversity and task relevance to promote effective and efficient collaboration and enhance overall system performance. Experiments show that AgentInit consistently outperforms state-of-the-art initialization methods and pre-defined strategies across various frameworks and tasks, achieving an overall performance improvement of up to 1.2 and 1.6, respectively, while also significantly reducing token consumption. Further analysis confirms its strong transferability to similar tasks and verifies the effectiveness of its key components, demonstrating its capability and adaptability as a reliable MAS initialization method. Source code and models are available at https://github.com/1737423697/AgentInit.

Federated learning in practice must contend with heterogeneous feature spaces, severe non-IID data, and scarce labels across clients. We present FedFusion, a federated transfer-learning framework that unifies domain adaptation and frugal labelling with diversity-/cluster-aware encoders (DivEn, DivEn-mix, DivEn-c). Labelled teacher clients guide learner clients via confidence-filtered pseudo-labels and domain-adaptive transfer, while clients maintain personalised encoders tailored to local data. To preserve global coherence under heterogeneity, FedFusion employs similarity-weighted classifier coupling (with optional cluster-wise averaging), mitigating dominance by data-rich sites and improving minority-client performance. The frugal-labelling pipeline combines self-/semi-supervised pretext training with selective fine-tuning, reducing annotation demands without sharing raw data. Across tabular and imaging benchmarks under IID, non-IID, and label-scarce regimes, FedFusion consistently outperforms state-of-the-art baselines in accuracy, robustness, and fairness while maintaining comparable communication and computation budgets. These results show that harmonising personalisation, domain adaptation, and label efficiency is an effective recipe for robust federated learning under real-world constraints.

Abstract--Federated Learning (FL) has emerged as a powerful paradigm for privacy-preserving model training, yet deployments in sensitive domains such as healthcare face persistent challenges from non-IID data, client unreliability, and adversarial manipulation. This paper introduces F edFiTS, a trust-and fairness-aware selective FL framework that advances the FedFaSt line by combining fitness-based client election with slotted aggregation. FedFiTS implements a three-phase participation strategy--free-for-all training, natural selection, and slotted team participation--augmented with dynamic client scoring, adaptive thresh-olding, and cohort-based scheduling to balance convergence efficiency with robustness. A theoretical convergence analysis establishes bounds for both convex and non-convex objectives under standard assumptions, while a communication-complexity analysis shows reductions relative to FedA vg and other baselines. Experiments on diverse datasets--medical imaging (X-ray pneumonia), vision benchmarks (MNIST, FMNIST), and tabular agricultural data (Crop Recommendation)--demonstrate that FedFiTS consistently outperforms FedA vg, FedRand, and FedPow in accuracy, time-to-target, and resilience to poisoning attacks. By integrating trust-aware aggregation with fairness-oriented client selection, FedFiTS advances scalable and secure FL, making it well suited for real-world healthcare and cross-domain deployments. The digitisation of healthcare and advances in artificial intelligence (AI) have reshaped medical data usage, enabling improved decision-making. Federated Learning (FL) represents a pivotal development, allowing collaborative model training across institutions without compromising data privacy, making it a crucial aspect in medical contexts. However, FL in healthcare must address challenges related to trust, transparency, security, and fairness.