Claire chatted to Chenying Liu from University of Oxford about how a robot's physical form can actively contribute to sensing, processing, decision-making, and movement. Chenying Liu is a Junior Research Fellow and an Associate Member of Faculty in the Department of Engineering Science at the University of Oxford. She leads an independent research programme focused on embodied physical intelligence, exploring how robot design can integrate geometry, materials, and control to enhance autonomy and robustness. Her work aims to develop more efficient and resilient robotic systems by embedding intelligence directly into their physical structures. Robot Talk is a weekly podcast that explores the exciting world of robotics, artificial intelligence and autonomous machines.

China's DeepSeek unveils latest models a year after upending global tech China's DeepSeek has unveiled the latest versions of its signature artificial intelligence-powered chatbot, a year after its flagship model sent shockwaves through the global tech scene. The Chinese start-up launched preview versions of DeepSeek-V4-Pro and DeepSeek-V4-Flash on Friday as it touted its ability to go toe-to-toe with US rivals such as OpenAI and Google. The "flash" model has similar reasoning abilities to the "pro" version, while offering faster response times and more cost-effective pricing, the Hangzhou-based startup said. Like DeepSeek's previous chatbots, V4-Pro and V4-Flash follow an open-source model, meaning developers are free to use and modify them at will. The release comes after DeepSeek-R1 stunned the tech sector upon its launch in January last year with capabilities broadly comparable with those of ChatGPT and Gemini.

In many application domains (e.g., recommender systems, intelligent tutoring systems), the rewards associated to the actions tend to decrease over time. This decay is either caused by the actions executed in the past (e.g., a user may get bored when songs of the same genre are recommended over and over) or by an external factor (e.g., content becomes outdated). These two situations can be modeled as specific instances of the rested and restless bandit settings, where arms are rotting (i.e., their value decrease over time). These problems were thought to be significantly different, since Levine et al. (2017) showed that state-of-the-art algorithms for restless bandit perform poorly in the rested rotting setting. In this paper, we introduce a novel algorithm, Rotting Adaptive Window UCB (RAW-UCB), that achieves near-optimal regret in both rotting rested and restless bandit, without any prior knowledge of the setting (rested or restless) and the type of non-stationarity (e.g., piece-wise constant, bounded variation). This is in striking contrast with previous negative results showing that no algorithm can achieve similar results as soon as rewards are allowed to increase. We confirm our theoretical findings on a number of synthetic and dataset-based experiments.

Diffusion-based generative models have reformed generative AI, and have enabled new capabilities in the science domain, for example, generating 3D structures of molecules. Due to the intrinsic problem structure of certain tasks, there is often a symmetry in the system, which identifies objects that can be converted by a group action as equivalent, hence the target distribution is essentially defined on the quotient space with respect to the group. In this work, we establish a formal framework for diffusion modeling on a general quotient space, and apply it to molecular structure generation which follows the special Euclidean group $\text{SE}(3)$ symmetry. The framework reduces the necessity of learning the component corresponding to the group action, hence simplifies learning difficulty over conventional group-equivariant diffusion models, and the sampler guarantees recovering the target distribution, while heuristic alignment strategies lack proper samplers. The arguments are empirically validated on structure generation for small molecules and proteins, indicating that the principled quotient-space diffusion model provides a new framework that outperforms previous symmetry treatments.

We introduce and study the problem of calibrating conditional risk, which involves estimating the expected loss of a prediction model conditional on input features. We analyze this problem in both classification and regression settings and show that it is fundamentally equivalent to a standard regression task. For classification settings, we further establish a connection between conditional risk calibration and individual/conditional probability calibration, and develop theoretical insights for the performance metric. This reveals that while conditional risk calibration is related to existing uncertainty quantification problems, it remains a distinct and standalone machine learning problem. Empirically, we validate our theoretical findings and demonstrate the practical implications of conditional risk calibration in the learning to defer (L2D) framework. Our systematic experiments provide both qualitative and quantitative assessments, offering guidance for future research in uncertainty-aware decision-making.

In this paper, it is shown, for the first time, that centralized performance is achievable in decentralized learning without sharing the local datasets. Specifically, when clients adopt an empirical risk minimization with relative-entropy regularization (ERM-RER) learning framework and a forward-backward communication between clients is established, it suffices to share the locally obtained Gibbs measures to achieve the same performance as that of a centralized ERM-RER with access to all the datasets. The core idea is that the Gibbs measure produced by client~$k$ is used, as reference measure, by client~$k+1$. This effectively establishes a principled way to encode prior information through a reference measure. In particular, achieving centralized performance in the decentralized setting requires a specific scaling of the regularization factors with the local sample sizes. Overall, this result opens the door to novel decentralized learning paradigms that shift the collaboration strategy from sharing data to sharing the local inductive bias via the reference measures over the set of models.

In an article published today in Nature, Sony AI introduce Ace, the first robot to beat elite human players in competitive physical sport. Although AI systems have shown advanced performance in digital domains and board games (such as complex video games, chess and Go), translating this to physical performance has remained a significant challenge. Such a feat requires perception, planning, and control to work in a high-speed domain on the scale of milliseconds. Table tennis is a demanding and complex real-world test for robotics, requiring rapid decision-making, precise physical execution, and continuous adaptation to an unpredictable opponent. The ball's high speed, spin, and complex trajectories are central to competitive play.

What Gilgo Beach killer's wife REALLY knew: Prosecutor reveals chilling truth about life with monster husband... and the'interests' she couldn't ignore Texas bride airlifted back to US on emergency flight after suffering'life-threatening' illness on honeymoon in Japan I thought I'd quit my addiction to'tweakments' and Botox forever. Then, feeling particularly confident at a Dubai lunch, I asked a stranger to guess my age... The lie my husband told to stop me divorcing him is beyond unforgivable. Every woman must beware... otherwise you might never realize: DEAR JANE Elizabeth Smart stuns fans with new incredible bodybuilding photos: 'I refuse to be ashamed' Dark secrets Days of Our Lives star Patrick Muldoon took to his grave: He'tried to hide' truth for decades... now friends are all whispering the same thing after his shock death New'Hollywood dose' pill: A-listers hooked on'youth elixir' that dermatologists say is anti-aging, shrinks pores, smooths wrinkles... and even banishes rosacea Katie Holmes likes telling comment about ex Joshua Jackson who shot to fame with her on Dawson's Creek Trump threatens to'blow up the rest of' Iran and'its leaders' with new Strait of Hormuz ultimatum'Paranoid' Tiger Woods and Vanessa Trump make major shakeup in the wake of golf legend's DUI scandal Death row inmate Chadwick Willacy who burned Florida mom alive during burglary is executed in front of victim's son What has Adam Levine done to his face? Meta's Big Brother move: Mark Zuckerberg's firm starts tracking employees' mouse clicks and taking screenshots of their screens - as one worker calls it'very dystopian' Meta has revealed plans to start tracking its employees' keystrokes and mouse clicks.

In the past, voice assistants like Siri or the one in your satnav used so-called'synthetic voices'. These require voice actors to spend hours in the recording studio, meticulously sampling all the different words and phrases that the assistant might need. Voice clones, on the other hand, have revolutionised how synthetic voices are created, by using AI to digitally recreate someone's speech patterns. These clones can be created with as little as a few seconds of recorded audio, even using clips from social media or snippets of conversation as the raw material. This has sparked concerns that criminals using AI could easily impersonate friends, family, or co-workers to manipulate their targets . According to the National Trading Standards, criminals are already using AI to clone people's voices and set up unauthorised direct debits over the phone. In the study, the researchers created voice clones of human participants using just 120 pre-recorded sentences. Participants listened to 80 unique sentences - 40 spoken by a real person and 40 spoken by an AI voice clone. The researchers compared human (top) AI-generated (bottom) voice recordings to see why this might be the case, but couldn't find any clear explanation Can you tell which voices are AI?

Inference-time LLM alignment methods, particularly activation steering, offer an alternative to fine-tuning by directly modifying activations during generation. Existing methods, however, often rely on non-anticipative interventions that ignore how perturbations propagate through transformer layers and lack online error feedback, resulting in suboptimal, open-loop control. To address this, we show empirically that, despite the nonlinear structure of transformer blocks, layer-wise dynamics across multiple LLM architectures and scales are well-approximated by locally-linear models. Exploiting this property, we model LLM inference as a linear time-varying dynamical system and adapt the classical linear quadratic regulator to compute feedback controllers using layer-wise Jacobians, steering activations toward desired semantic setpoints in closed-loop with minimal computational overhead and no offline training. We also derive theoretical bounds on setpoint tracking error, enabling formal guarantees on steering performance. Using a novel adaptive semantic feature setpoint signal, our method yields robust, fine-grained behavior control across models, scales, and tasks, including state-of-the-art modulation of toxicity, truthfulness, refusal, and arbitrary concepts, surpassing baseline steering methods. Our code is available at: https://github.com/trustworthyrobotics/lqr-activation-steering