Unmasking Exoplanets: AI Filters Out Stellar Noise
New research demonstrates that artificial intelligence can effectively remove the ‘jitter’ caused by star activity, improving our ability to find Earth-like planets.
Science
Learning to See Beyond the Seen: Robust Vision with Equivariant Operators
New research demonstrates how deep neural networks can achieve greater robustness and generalization in image recognition by learning underlying transformations, even with limited data.
Learning to Spin: Machine Learning Accelerates Magnetic Dynamics
![The model utilizes a machine learning force field to map complex spin configurations to local energies [latex]\epsilon_i = \varepsilon(\mathcal{C}_i)[/latex], enabling the computation of total potential energy through summation and, crucially, leveraging automatic differentiation to derive local exchange fields [latex]\mathbf{H}_i[/latex] from the energy’s derivatives with respect to individual spins [latex]\partial E/\partial\mathbf{S}_i[/latex].](https://arxiv.org/html/2602.18213v1/x1.png)
A new machine learning framework dramatically speeds up simulations of how magnetism evolves in metals, opening doors to modeling complex magnetic phenomena.
Smarter AI: Learning from Data Without Constant Trial and Error
New algorithms are pushing the boundaries of offline reinforcement learning, enabling AI agents to learn optimal policies from static datasets and minimizing the need for costly real-world interactions.
GNNs Meet Hard Problems: A Reality Check for Constraint Solving
![NeuroSAT, a Graph Neural Network, demonstrates that scaling message-passing iterations linearly with problem size [latex]\alpha = M/N[/latex] yields significantly improved performance in finding satisfying assignments for 3-SAT problems, consistently outperforming both supervised training and fixed-iteration approaches-a result achieved by maintaining consistent inference time regardless of problem scale.](https://arxiv.org/html/2602.18419v1/x6.png)
A new benchmark reveals the current limitations of graph neural networks when tackling complex constraint satisfaction problems, demonstrating they lag behind traditional algorithms at scale.
Sharpening AI Minds: How Challenging Questions Boost Specialized Reasoning
A new framework uses adversarial question generation to push the limits of smaller language models in complex, domain-specific tasks.
Beyond Familiar Paths: Steering Generative Models with Divergent Trajectories
A new approach balances exploration and exploitation in generative flow networks, enabling more efficient learning and improved sample quality.
Lost in the Documents: Fixing Retrieval for Financial AI
Accurate question answering over long financial reports hinges on effective information retrieval, but current systems often struggle to find the right data.
When AI Gets It Wrong: How Students Spot—and Struggle With—Machine Hallucinations
A new study examines university students’ understanding of artificial intelligence ‘hallucinations’ – instances where AI confidently presents false information – and their ability to identify and address these errors.
Decoding Galaxy Spectra with Deep Learning
New research leverages unsupervised neural networks to analyze the complex spectral fingerprints of galaxies, paving the way for the discovery of unusual celestial objects.