The BBC has upheld 20 complaints over impartiality after presenter Martine Croxall altered a script she was reading live on the BBC News Channel which referred to pregnant people earlier this year. Croxall was introducing an interview about research on groups most at risk during UK heatwaves, which quoted a release from the London School of Hygiene and Tropical Medicine. The presenter changed her script to instead say women, and the BBC's Executive Complaints Unit said it considered her facial expression to express a controverial view about trans people. The presenter said: Malcolm Mistry, who was involved in the research, says that the aged, pregnant people women and those with pre-existing health conditions need to take precautions. The ECU said it considered Croxall's facial expression laid it open to the interpretation that it indicated a particular viewpoint in the controversies currently surrounding trans ideology.

Six new cookbooks bring stellar dishes--and cultures--from around the world into your kitchen. All products featured on WIRED are independently selected by our editors. However, we may receive compensation from retailers and/or from purchases of products through these links. Think about how difficult cooking from a cookbook from another culture was as little as 10 years ago. Once in a while, you could get your hands on a standout, but the food you could make with it could feel like a compromise with too many substitutions and ingredients you just couldn't find without great effort, or at all.

The AI Song Contest was founded with the aim of showcasing the potential of human-AI co-creativity in the songwriting process. Now in its sixth year, the competition will conclude on 16 November with a live show in Amsterdam. From all the entrants, the jury have selected their top ten songs. The live event will feature performances from the ten finalists, and you will be able to watch on YouTube here . Listen to the songs and vote for your favourite.

Is Trump losing patience with Putin? Will sanctions against Russian oil giants hurt Putin? Russian forces have spread rapidly through Pokrovsk, the city in Ukraine's east where the warring sides have concentrated their manpower and tactical ingenuity during the past week, in what may be a final culmination of a 21-month battle. Geolocated footage placed Russian troops in central, northern and northeastern Pokrovsk, said the Institute for the Study of War (ISW), a Washington-based think tank. It set its sights on the city almost two years ago, after capturing Avdiivka, 39km (24 miles) to the east.

'Vibe coding' named word of the year by Collins Dictionary If you've ever wanted to create your own computer program but never learnt how to code, you might try vibe coding. Collins Dictionary's word of the year - which is confusingly made up of two words - is the art of making an app or website by describing it to artificial intelligence (AI) rather than by writing programming code manually. The term was coined in February by OpenAI co-founder Andrej Karpathy, who came up with the name to represent how AI can let some programmers forget that the code even exists and give in to the vibes while making a computer program. It was one of 10 words on a shortlist to reflect the mood, language and preoccupations of 2025. By giving an AI tool a simple description such as make me a program that schedules my weekly meals, people can use vibe coding to make basic apps without any previous programming knowledge.

From gas to groceries, has Trump kept his promise to tackle rising prices? President Donald Trump was swept to power for a second time on the back of a central campaign promise to tackle inflation. The steep rise in the cost of living was top of voters' minds and Trump blamed President Joe Biden. He also made sweeping promises to bring down prices for Americans starting on day one. One year on from his victory, BBC Verify revisits some of the president's claims.

Will quantum be bigger than AI? There's an old adage among tech journalists like me - you can either explain quantum accurately, or in a way that people understand, but you can't do both. That's because quantum mechanics - a strange and partly theoretical branch of physics - is a fiendishly difficult concept to get your head around. It involves tiny particles behaving in weird ways. And this odd activity has opened up the potential of a whole new world of scientific super power. Its mind-boggling complexity is probably a factor in why quantum has ended up with a lower profile than tech's current rockstar - artificial intelligence (AI).

Large language models (LLMs) excel at complex reasoning tasks but remain computationally expensive, limiting their practical deployment. To address this, recent works have focused on distilling reasoning capabilities into smaller language models (sLMs) using chain-of-thought (CoT) traces from teacher LLMs. However, this approach struggles in scenarios requiring rare factual knowledge or precise computation, where sLMs often hallucinate due to limited capability. In this work, we propose Agent Distillation, a framework for transferring not only reasoning capability but full task-solving behavior from LLM-based agents into sLMs with retrieval and code tools. We improve agent distillation along two complementary axes: (1) we introduce a prompting method called first-thought prefix to enhance the quality of teacher-generated trajectories; and (2) we propose a self-consistent action generation for improving test-time robustness of small agents. We evaluate our method on eight reasoning tasks across factual and mathematical domains, covering both in-domain and out-of-domain generalization. Our results show that sLMs as small as 0.5B, 1.5B, 3B parameters can achieve performance competitive with next-tier larger 1.5B, 3B, 7B models fine-tuned using CoT distillation, demonstrating the potential of agent distillation for building practical, tool-using small agents. Our code is available at https://github.com/Nardien/agent-distillation.

In this research, we present SLYKLatent, a novel approach for enhancing gaze estimation by addressing appearance instability challenges in datasets due to aleatoric uncertainties, covariant shifts, and test domain generalization. SLYKLatent utilizes Self-Supervised Learning for initial training with facial expression datasets, followed by refinement with a patch-based tri-branch network and an inverse explained variance-weighted training loss function. Our evaluation on benchmark datasets achieves a 10.9% improvement on Gaze360, supersedes top MPIIFaceGaze results with 3.8%, and leads on a subset of ETH-XGaze by 11.6%, surpassing existing methods by significant margins. Adaptability tests on RAF-DB and Affectnet show 86.4% and 60.9% accuracies, respectively. Ablation studies confirm the effectiveness of SLYKLatent's novel components.

This paper presents ECGXtract, a deep learning-based approach for interpretable ECG feature extraction, addressing the limitations of traditional signal processing and black-box machine learning methods. In particular, we develop convolutional neural network models capable of extracting both temporal and morphological features with strong correlations to a clinically validated ground truth. Initially, each model is trained to extract a single feature, ensuring precise and interpretable outputs. A series of experiments is then carried out to evaluate the proposed method across multiple setups, including global versus lead-specific features, different sampling frequencies, and comparisons with other approaches such as ECGdeli. Our findings show that ECGXtract achieves robust performance across most features with a mean correlation score of 0.80 with the ground truth for global features, with lead II consistently providing the best results. For lead-specific features, ECGXtract achieves a mean correlation score of 0.822. Moreover, ECGXtract achieves superior results to the state-of-the-art open source ECGdeli as it got a higher correlation score with the ground truth in 90% of the features. Furthermore, we explore the feasibility of extracting multiple features simultaneously utilizing a single model. Semantic grouping is proved to be effective for global features, while large-scale grouping and lead-specific multi-output models show notable performance drops. These results highlight the potential of structured grouping strategies to balance the computational efficiency vs. model accuracy, paving the way for more scalable and clinically interpretable ECG feature extraction systems in limited resource settings.