We study the sample complexity of robust binary hypothesis testing under three standard contamination models: $\varepsilon$-additive (Huber), $\varepsilon$-subtractive, and $\varepsilon$-total variation (TV), denoted by $n^*_{\mathrm{Hub}}(\varepsilon)$, $n^*_{\mathrm{Sub}}(\varepsilon)$, and $n^*_{\mathrm{TV}}(\varepsilon)$, respectively. For subtractive contamination, we show that least favourable distributions exist and provide explicit formulas for the same, bringing this model in line with the classical Huber and TV models. Next we show that in all three models, sample complexity may be highly unstable in the contamination parameter $\varepsilon$, increasing by polynomial factors even for $o(\varepsilon)$ perturbations. Similarly, there may be polynomial factor gaps between the sample complexities when $\varepsilon$ is known exactly versus when it is known up to $o(\varepsilon)$ error. Despite the instability of the sample complexity in all models, we show that the sample complexities across models are comparable up to constant-factor rescaling of $\varepsilon$. Specifically, for any fixed $δ_0>0$, the following hold for all distributions $p$ and $q$: (i) $n^*_{\mathrm{Hub}}(\varepsilon) \lesssim n^*_{\mathrm{TV}}(\varepsilon) \lesssim n^*_{\mathrm{Hub}}(2\varepsilon)$, (ii) $n^*_{\mathrm{Sub}}(\varepsilon) \lesssim n^*_{\mathrm{TV}}(\varepsilon) \lesssim n^*_{\mathrm{Sub}}((2+δ_0)\varepsilon)$, and (iii) $n^*_{\mathrm{Sub}}(\varepsilon) \lesssim n^*_{\mathrm{Hub}}(\varepsilon) \lesssim n^*_{\mathrm{Sub}}((1+δ_0)\varepsilon)$, and the scaling constants are tight. Finally, we extend our results to adaptive versions of the contamination models.

Spectral deconvolution is essential for extracting peak structures that encode material properties and chemical structures, but conventional automated methods often fail when spectra contain high-intensity noise or unknown background components. In practice, scientists rarely interpret spectra in isolation. Instead, they identify physically meaningful peaks by relating spectral structures to auxiliary information such as physical-property values, chemical structures, and trends across related measurements. Here, we propose a Bayesian framework that integrates spectral deconvolution with a model of expert scientific reasoning. In this work, expert scientific reasoning refers to the practice of evaluating candidate spectral structures by their consistency with independently measured physical-property values, rather than to manual expert intervention during inference. We formalize this reasoning as a physical-property regression layer, implemented using Gaussian process regression, and couple it with Bayesian spectral deconvolution. By averaging the physical-property likelihood over posterior predictive spectra inferred from Bayesian spectral deconvolution, the proposed method selects spectral models according to the consistency between inferred spectral structures and physical-property information. We validate the framework using synthetic spectra with high-intensity noise or unknown backgrounds and infrared spectra of poly(lactic acid). The method recovers physically meaningful peak structures that conventional Bayesian spectral deconvolution misses or misidentifies from spectra alone, including weak peaks in poly(lactic acid) IR spectra related to measured degradation rates. These results demonstrate that integrating expert scientific reasoning with Bayesian spectral deconvolution enables robust peak estimation under conditions where spectrum-only inference is unreliable.

On the 31st March, our editorial team headed to the Royal Society for AI for Science . This day-long conference explored how AI is changing the nature of scientific discovery, and was hosted by the Fundamental Research team from the Alan Turing Institute. Nestled in a terrace of 19th century townhouses along the banks of the Thames, the Royal Society looks as grand as the names who have passed through its doors throughout the years. Prof Jason McEwen, Chief Scientist for the Turing Institute, opened the event with an insightful talk on the nature of scientific revolution, and how the bidirectional relationship between AI and science could spark the next one. Then, Prof Anna Scaife from the University of Manchester spoke on the use of foundation models for astronomical discovery.

If the goal of a robot is to perform a function, then minimizing the possibility of failure is a top priority when it comes to robotic design. But this minimization is at odds with the robotic raison d'être: systems with multiple units, or agents, can perform more diverse functions, but they also have more different parts that can potentially fail. Researchers led by Jamie Paik, head of the Reconfigurable Robotics Laboratory ( RRL) in EPFL's School of Engineering, have not only circumvented this problem, but flipped it: they have designed a modular robot that actually lowers its odds of failure by sharing resources among its individual agents. "For the first time, we have found a way to reverse the trend of increasing odds of failure with increasing function," Paik explains. "We introduce local resource sharing as a new paradigm in robotics, reducing the failure rate with a larger number of modules."

What I've learned from 25 years of automated science, and what the future holds: an interview with Ross King We're excited to launch our new series, where we're speaking with leading researchers to explore the breakthroughs driving AI and the reality of the future promises - to give you an inside perspective on the headlines. Our first interviewee is Ross King, who created the first robot scientist back in 2009. He spoke to us about the nature of scientific discovery, the role AI has to play, and his recent work in DNA computing. Automated science is a really exciting area, and it feels like everyone's talking about it at the moment - e.g. But you've been working in this field for many years now. In 2009 you developed Adam, the first robot scientist to generate novel scientific knowledge. Could you tell me some more about that? So the history goes back to before Adam.

Aligning human-interpretable concepts with the internal representations learned by modern machine learning systems remains a central challenge for interpretable AI. We introduce a geometric framework for comparing supervised human concepts with unsupervised intermediate representations extracted from foundation model embeddings. Motivated by the role of conceptual leaps in scientific discovery, we formalise the notion of concept frustration: a contradiction that arises when an unobserved concept induces relationships between known concepts that cannot be made consistent within an existing ontology. We develop task-aligned similarity measures that detect concept frustration between supervised concept-based models and unsupervised representations derived from foundation models, and show that the phenomenon is detectable in task-aligned geometry while conventional Euclidean comparisons fail. Under a linear-Gaussian generative model we derive a closed-form expression for Bayes-optimal concept-based classifier accuracy, decomposing predictive signal into known-known, known-unknown and unknown-unknown contributions and identifying analytically where frustration affects performance. Experiments on synthetic data and real language and vision tasks demonstrate that frustration can be detected in foundation model representations and that incorporating a frustrating concept into an interpretable model reorganises the geometry of learned concept representations, to better align human and machine reasoning. These results suggest a principled framework for diagnosing incomplete concept ontologies and aligning human and machine conceptual reasoning, with implications for the development and validation of safe interpretable AI for high-risk applications.

Automated scientific discovery promises to accelerate progress across scientific domains, but evaluating an agent's capacity for end-to-end scientific reasoning is challenging as running real-world experiments is often prohibitively expensive or infeasible. In this work we introduce DiscoveryWorld, a virtual environment that enables benchmarking an agent's ability to perform complete cycles of novel scientific discovery in an inexpensive, simulated, multi-modal, long-horizon, and fictional setting.DiscoveryWorld consists of 24 scientific tasks across three levels of difficulty, each with parametric variations that provide new discoveries for agents to make across runs. Tasks require an agent to form hypotheses, design and run experiments, analyze results, and act on conclusions. Task difficulties are normed to range from straightforward to challenging for human scientists with advanced degrees. DiscoveryWorld further provides three automatic metrics for evaluating performance, including: (1) binary task completion, (2) fine-grained report cards detailing procedural scoring of task-relevant actions, and (3) the accuracy of discovered explanatory knowledge.While simulated environments such as DiscoveryWorld are low-fidelity compared to the real world, we find that strong baseline agents struggle on most DiscoveryWorld tasks, highlighting the utility of using simulated environments as proxy tasks for near-term development of scientific discovery competency in agents.

Neural Information Processing Systems http://nips.cc/

Neural Information Processing Systems http://nips.cc/