Complex games have long been an important benchmark for testing the progress of artificial intelligence algorithms. AlphaGo, AlphaZero, and MuZero have defeated top human players in Go and Chess, garnering widespread societal attention towards artificial intelligence. Concurrently, large language models (LLMs) have exhibited remarkable capabilities across various tasks, raising the question of whether LLMs can achieve similar success in complex games. In this paper, we explore the potential of LLMs in mastering complex card games. We systematically assess the learning capabilities of LLMs across eight diverse card games, evaluating the impact of fine-tuning on high-quality gameplay data, and examining the models' ability to retain general capabilities while mastering these games. Our findings indicate that: (1) LLMs can approach the performance of strong game AIs through supervised fine-tuning on high-quality data, (2) LLMs can achieve a certain level of proficiency in multiple complex card games simultaneously, with performance augmentation for games with similar rules and conflicts for dissimilar ones, and (3) LLMs experience a decline in general capabilities when mastering complex games, but this decline can be mitigated by integrating a certain amount of general instruction data. The evaluation results demonstrate strong learning ability and versatility of LLMs.

Many real-world problems require reasoning across multiple scales, demanding models which operate not on single data points, but on entire distributions. We introduce generative distribution embeddings (GDE), a framework that lifts autoencoders to the space of distributions. In GDEs, an encoder acts on sets of samples, and the decoder is replaced by a generator which aims to match the input distribution. This framework enables learning representations of distributions by coupling conditional generative models with encoder networks which satisfy a criterion we call distributional invariance. We show that GDEs learn predictive sufficient statistics embedded in the Wasserstein space, such that latent GDE distances approximately recover the $W_2$ distance, and latent interpolation approximately recovers optimal transport trajectories for Gaussian and Gaussian mixture distributions.

Bang Chun-ja, a 78-year-old South Korean woman living alone, holding Hyodol, an artificial intelligence-powered healthcare doll designed for the elderly, during an interview at her home in Yongin in April. Yongin, South Korea - In her tiny apartment in South Korea where she lives alone, 78-year-old Bang Chun-ja spends her days with a childlike artificial intelligence-powered doll she says she prefers to people. The doll greets Bang when she returns home, sings to her when she feels bored, reminds her not to skip meals or medication -- helping her maintain a routine -- and tells her it loves her. Bang has limited contact with her grown-up daughter, and fell into severe depression after major back surgery, spending hours alone staring at the ceiling in pain. In a time of both misinformation and too much information, quality journalism is more crucial than ever.

Reinforcement learning (RL) systems have countless applications, from energy-grid management to protein design. However, such real-world scenarios are often extremely difficult, combinatorial in nature, and require complex coordination between multiple agents. This level of complexity can cause even state-of-the-art RL systems, trained until convergence, to hit a performance ceiling which they are unable to break out of with zero-shot inference. Meanwhile, many digital or simulation-based applications allow for an inference phase that utilises a specific time and compute budget to explore multiple attempts before outputting a final solution. In this work, we show that such an inference phase employed at execution time, and the choice of a corresponding inference strategy, are key to breaking the performance ceiling observed in complex multi-agent RL problems. Our main result is striking: we can obtain up to a 126% and, on average, a 45% improvement over the previous state-of-the-art across 17 tasks, using only a couple seconds of extra wall-clock time during execution. We also demonstrate promising compute scaling properties, supported by over 60k experiments, making it the largest study on inference strategies for complex RL to date. We make all of our experimental data and code available.

Reinforcement Learning (RL) traditionally relies on scalar reward signals, limiting its ability to leverage the rich semantic knowledge often available in real-world tasks. In contrast, humans learn efficiently by combining numerical feedback with language, prior knowledge, and common sense. We introduce Prompted Policy Search (ProPS), a novel RL method that unifies numerical and linguistic reasoning within a single framework. Unlike prior work that augments existing RL components with language, ProPS places a large language model (LLM) at the center of the policy optimization loop--directly proposing policy updates based on both reward feedback and natural language input. We show that LLMs can perform numerical optimization in-context, and that incorporating semantic signals, such as goals, constraints, and strategy hints can lead to more informed exploration and sample-efficient learning. ProPS is evaluated across 15 Gymnasium tasks, spanning classic control, Atari games, and MuJoCo environments, and compared to seven widely-adopted RL algorithms (e.g., PPO, SAC, TRPO). It outperforms all baselines on 8 out of 15 tasks and demonstrates substantial gains when provided with domain knowledge.

China-based actors are likely behind the use of ChatGPT for "covert influence operations" aimed at stoking opposition to data centres in the United States, OpenAI has said. In a research report released on Wednesday, the company behind the world's most popular AI chatbot said it had banned a cluster of accounts likely based in China for attempting to "manipulate a legitimate debate about American AI". Among other content, the accounts generated a comic strip showing a cigar-chomping businessman holding bags marked with dollar signs as a family reacted in shock to their electricity bill, according to the San Francisco-based company. OpenAI said a second cluster of accounts had generated content casting US tariffs as an effort to "dominate technological competition" with China, and specified that the material should not mention Chinese leader Xi Jinping. While the campaign sought to "exploit and amplify existing public concerns" about energy prices, OpenAI found no evidence that it had a "meaningful" influence, the company said.

A group of high school students arrested over allegedly trying to extort money from a boy in western Tokyo may have used ChatGPT to decide how much to demand, police said. A group of high school students in Tokyo arrested over allegedly assaulting a boy and trying to extort money from him may have used ChatGPT to decide how much to demand, media reports have recently revealed. Five teenagers, including a 17-year-old girl and four boys ranging in age from 16 to 17, were arrested in January over the alleged assault and attempted extortion of a 17-year-old high school student in the city of Hachioji in western Tokyo, according to the Metropolitan Police Department. Police said the suspects assaulted the boy in a plaza in Hachioji's Shiroyamate district, breaking his nose and causing other injuries, before allegedly trying to extort ¥150,000 ($935) from him. The girl, who was the victim's ex-girlfriend, allegedly first confronted him, accusing him of touching her younger sister's leg. She then challenged him, saying, "Give me the money or fight me one-on-one," according to reports by Fuji TV.

Anthropic Walks Back Policy That Could Have'Sabotaged' AI Researchers Using Claude The company changed course after researchers spoke out against the policy, which would have covertly limited Claude's ability to develop competing AI models. Anthropic is backtracking on a policy that would have covertly limited competitors from using its new AI model, Claude Fable 5, to develop other AI models. The company changed course after the move received significant backlash from the AI research community . "We're changing Fable 5's safeguards for frontier LLM development to make them visible," Anthropic said in a statement to WIRED. "We made the wrong tradeoff and we apologize for not getting the balance right."

Generative large language models (LLMs) have achieved state-of-the-art results on a wide range of tasks, yet they remain susceptible to backdoor attacks: carefully crafted triggers in the input can manipulate the model to produce adversary-specified outputs. While prior research has predominantly focused on backdoor risks in vision and classification settings, the vulnerability of LLMs in open-ended text generation remains underexplored.

Consistency learning with feature perturbation is a widely used strategy in semi-supervised medical image segmentation. However, many existing perturbation methods rely on dropout, and thus require a careful manual tuning of the dropout rate, which is a sensitive hyperparameter and often difficult to optimize and may lead to suboptimal regularization. To overcome this limitation, we propose VQ-Seg, the first approach to employ vector quantization (VQ) to discretize the feature space and introduce a novel and controllable Quantized Perturbation Module (QPM) that replaces dropout.