Russia and Ukraine exchanged 205 prisoners of war on Friday, hours after rescue workers ended their search of a destroyed block of flats in Kyiv in which 24 people were killed, including three girls. Most of the Ukrainian prisoners had been held since 2022, said President Zelensky. The swap was part of a short-lived ceasefire ending this week with the launch of massive Russian strikes across Ukraine, including a missile attack that reduced 18 flats to rubble. Among the victims was 12-year-old Lyubava Yakovleva, whose father was killed during the war. Meanwhile, Russian officials said four people, including a child, were killed when Ukrainian drones hit the city of Ryazan, south-east of Moscow.

Plus: as their trial goes to the jury, Musk and Altman face lying accusations. China's short drama industry is fueled by bite-sized, melodramatic, and smutty shows built for smartphone scrolling. Now, many are being made entirely with AI: no actors, camera operators, cinematographers, or CGI specialists required. An average of 470 AI-generated short dramas were released every day in January. Production timelines have shrunk from months to weeks, while costs have dropped by up to 90%. Storytelling is also increasingly driven by performance data.

CFTC chairman Michael Selig sat down with WIRED to discuss how the agency scours Polymarket and other prediction markets for illegal activity. For most of the past year, it looked like prediction markets had kicked off a new golden age of fraud. On Polymarket, traders raked in fortunes from suspiciously timed bets on geopolitical events like the raid on Venezuela and the Iran War. It wasn't clear whether the US government would bother pursuing some of the most flagrant bad actors, since Polymarket's crypto-based platform was technically offshore and not regulated or licensed within the country. Now, however, the Commodity Futures Trading Commission, which oversees prediction markets, wants you to know that it's watching very, very closely.

The viral short dramas are increasingly being created entirely with AI, with hundreds of new shows spun up each day. In a dimly lit bedroom, a frightened young woman is thrown onto a bed by a tall, muscular man. He grabs her hand, and flame-like vines crawl across her body, fusing with her flesh. A dragon-shaped tattoo appears across her chest. "Two months," the man says. "Give me an heir, or I will eat you."

Students at one of Russia's leading engineering universities are getting a lucrative offer: ditch their studies for a year, fly drones for the military and earn more than 5 million rubles ($68,275) in pay as well as free tuition on their return. Pamphlets distributed at Bauman Moscow State Technical University promise students who sign up for the unmanned systems forces will fly drones from far behind the front lines, but still qualify for combat veteran status. It's part of a broader push across Russia to recruit university and college students, using lavish signing bonuses, academic leave and even outright coercion to convince young men to join the fight. At least 270 institutions are actively promoting military contracts, according to the independent magazine Groza, which specializes in higher education and student issues. In a time of both misinformation and too much information, quality journalism is more crucial than ever. By subscribing, you can help us get the story right.

Before arriving for his high-stakes summit with Chinese leader Xi Jinping, United States President Donald Trump aimed to set expectations high. He said he would urge Xi to "open up" China's economy and announced a delegation of top business executives, including Tesla's Elon Musk, Apple's Tim Cook and Nvidia's Jensen Huang, to accompany him. While Trump and Xi are anticipated to extend the one-year pause in their trade war agreed to in South Korea in October, the expectations are for a stabilisation - not revitalisation - in ties between the world's two largest economies, which are locked in a rivalry that spans everything from trade and artificial intelligence to the status of Taiwan. "It is important to be clear-eyed about the state of relations here," Claire E Reade, a senior counsel at Arnold & Porter who previously worked on China at the Office of the US Trade Representative (USTR), told Al Jazeera. "China does not trust the US, and China wants to beat the US in what it sees as long-term global competition," Reade said.

Brutal raid on woman's birthday party highlights rise of Russian vigilante group Katya was about to blow out the candles on her 30th birthday cake when masked men burst into the nightclub hired for her party, and began physically and verbally attacking her friends. They called us faggots and lesbians. I could hear violence from every corner, she told a BBC World Service investigation. Her mother was told to get down on all fours, she says. The swoop was instigated by a vigilante group, called Russkaya Obshina, that wants to accelerate President Vladimir Putin's agenda to stamp out what he describes as Western liberalism, and promote traditional family-oriented values.

Physics-informed neural networks (PINNs) provide a mesh-free framework for solving PDE-constrained inverse problems, but their extension to Bayesian inversion still faces a fundamental difficulty: prior distributions are typically defined in the weight space of neural networks, whereas physically meaningful prior assumptions are more naturally expressed in function space. In this study, we introduce a unified framework, termed functional-prior-based approaches to Bayesian PDE-constrained inversion using physics-informed neural networks (fpBPINN), to incorporate functional priors into Bayesian PINN-based inversion. We consider two complementary approaches. The first is a functional-prior-informed Bayesian PINN (FPI-BPINN), in which a neural network weight prior is learned to be consistent with a prescribed functional prior, and Bayesian inference is subsequently performed in weight space. The second is function-space particle-based variational inference for PINNs (fParVI-PINN), which performs Bayesian estimation using ParVI directly in function space. We also show that random Fourier features (RFF) play an important role in representing Gaussian functional priors with neural networks and in improving posterior approximation. We applied the proposed approaches to one-dimensional seismic traveltime tomography and two-dimensional Darcy-flow permeability inversion. These numerical experiments showed that both approaches accurately estimated posterior distributions, highlighting the significance of introducing physically interpretable functional priors into Bayesian PINN-based inverse problems. We also identified the contrasting advantages of FPI-BPINN and fParVI-PINN, namely flexibility and accuracy, respectively.

Quantum machine learning (QML) aims to accelerate machine learning tasks by exploiting quantum computation. Previous work studied a QML algorithm for selecting sparse subnetworks from large shallow neural networks. Instead of directly solving an optimization problem over a large-scale network, this algorithm constructs a sparse subnetwork by sampling hidden nodes from an optimized probability distribution defined using the ridgelet transform. The quantum algorithm performs this sampling in time $O(D)$ in the data dimension $D$, whereas a naive classical implementation relies on handling exponentially many candidate nodes and hence takes $\exp[O(D)]$ time. In this work, we construct and analyze a quantum-inspired fully classical algorithm for the same sampling task. We show that our algorithm runs in time $O(\operatorname{poly}(D))$, thereby removing the exponential dependence on $D$ from the previous classical approach. Numerical simulations show that the proposed sampler achieves empirical risk comparable to exact sampling from the optimized distribution and substantially lower than sampling from the non-optimized uniform distribution, while also exhibiting exponentially improved runtime scaling compared with the conventional classical implementation. These successful dequantization results show that sparse subnetwork selection via optimized sampling can be achieved classically with polynomial data-dimension scaling on conventional computers without quantum hardware, providing an alternative to the existing quantum algorithm.

Tabular data underpins most high-value prediction problems in science and industry, and TabPFN has driven the foundation model revolution for this modality. Designed with feedback from our users, TabPFN-3 builds on this foundation to scale state-of-the-art performance to datasets with 1M training rows and substantially reduce training and inference time. Pretrained exclusively on synthetic data from our prior, TabPFN-3 dramatically pushes the frontier of tabular prediction and brings substantial gains on time series, relational, and tabular-text data. On the standard tabular benchmark TabArena, a forward pass of TabPFN-3 outperforms all other models, including tuned and ensembled baselines, by a significant margin, and pareto-dominates the speed/performance frontier. On more diverse datasets, TabPFN-3 ranks first on datasets with many classes, and beats 8-hour-tuned gradient-boosted-tree baselines on datasets up to 1M training rows and 200 features. TabPFN-3 introduces test-time compute scaling to tabular foundation models. Our API offering TabPFN-3-Plus (Thinking) exploits this to beat all non-TabPFN models by over 200 Elo on TabArena, rising to 420 Elo on the largest data subset, and outperforms AutoGluon 1.5 extreme while being 10x faster, without using LLMs, real data, internet search or any other model besides TabPFN. TabPFN-3 extends the capabilities of our models, enabling SOTA prediction on relational data (new SOTA foundation model on RelBenchV1) and tabular-text data (SOTA on TabSTAR via TabPFN-3-Plus); and improves existing integrations: a specialized checkpoint, TabPFN-TS-3, ranks 2nd on the time-series benchmark fev-bench, and SHAP-value computation is up to 120x faster. TabPFN-3 achieves this performance while being up to 20x faster than TabPFN-2.5. In addition, a reduced KV cache and row-chunking scale to 1M rows on one H100 with fast inference speed.