scientific-computing

The Emscripten-forge NumPy package now links OpenBLAS in WebAssembly, delivering up to 30x faster matrix operations in the browser for float32.

This paper evaluates AI's performance against traditional statistics and scientific computing in 27 scientific disciplines, finding that AI often outperforms statistics at higher computational cost but increasingly outperforms computing at lower cost, reshaping the scientific frontier.

This paper introduces topology as a distinct generalization axis for neural operators in PDE solving, using Hodge heat flow and the TopoBox-3D framework to show its influence across the Hodge spectrum beyond geometric compatibility.

HarmoCore introduces a functional latent diffusion method for reconstructing oscillatory wave fields from sparse sensor data, achieving significant improvements in efficiency and accuracy with minimal sensing in 2D and 3D scenarios.

The paper proposes GeoLAMP, a geometry-aware latent autoregressive generative model for solving multiphysics partial differential equations in complex geometries, using a dual-encoder architecture and causal self-attention transformer with flow matching for stable and scalable predictions.

The article traces the origins of the Julia programming language from an MIT research project to a global tool used by over a million people, and highlights the launch of Dyad 3.0, an AI platform by JuliaHub for automating engineering simulations.

This paper proposes a Physics-Informed Error Field Learning (PIEFL) framework for Physics-Informed Neural Networks, introducing an auxiliary error network to improve solution accuracy in solving partial differential equations under computational constraints.

This paper introduces a physics-integrated operator learning framework using a feed-forward Gaussian splatting representation to directly incorporate PDE operators, reducing errors in long-horizon autoregressive predictions for spatiotemporal systems.

This paper presents a method to rank neural operator models during deployment using shared physics responses, achieving high accuracy without ground-truth reference solutions for scientific computing applications.

The article introduces a rationally enriched Chebyshev trunk for DeepONet surrogate models, enhancing accuracy in simulating high-Péclet transport problems with thin boundary layers.

This study systematically evaluates physics-informed neural network techniques for fluid dynamics, showing that combining periodic activations with causal weighting improves performance on the Navier-Stokes vortex shedding benchmark, while further additions cause degradation due to nonlinear interactions.

This paper proposes Fourier Feature Networks (FENs), a single-hidden-layer neural network architecture using Fourier features to solve partial differential equations, achieving higher accuracy than Extreme Learning Machines without affine transformations.

PIKFNO is a new interpretable neural operator framework that integrates physics-informed kernel functions from governing equations to enhance predictive accuracy and interpretability with limited training data.

RECAST is a machine-learning framework that combines learned correction and super-resolution to restore accuracy in coarse-grid PDE solvers, reducing error by 50-92% across six 1D PDE systems and enabling coarser simulations without losing fidelity.

Basin is a numerical optimization library for Rust that provides a broad catalog of solvers, first-class constraint support, and a WebAssembly-ready default build.

This paper introduces Distribird, an agentic web application that automates the design of informative prior distributions for Bayesian model calibration by searching and reading scientific literature. It evaluates the tool on 24 parameters across ten domains using local open-weight LLMs, emphasizing traceability, privacy, and out-of-scope rejection over raw accuracy.

Agent-MD is a framework that selectively applies LLM reasoning to long-running molecular simulation campaigns, using a deterministic rule-based agent for routine tasks and event-triggered LLM review for exceptional conditions. Demonstrated in GCMC–MD water-vapor desorption simulations, it shows that auditable, reproducible scientific workflows can avoid placing every operation inside an LLM reasoning loop.

A software engineer reflects on how scientists often lack software engineering skills, using an example of optimizing an astrophysics simulation postprocessing tool and advocating for a 'missing semester' for scientists.

EvoPINN is an agentic framework that reformulates PINN development as an execution-grounded algorithm discovery problem, using an LLM agent to propose programmatic modifications. It autonomously discovers PDE-specialized learning algorithms, including a novel architecture called SLRC-PINN, which outperforms baselines across diverse PDE regimes.

OpenAI highlights how coding agents can assist scientists with research tasks such as routine maintenance and optimization, while emphasizing that researchers must still define scientific questions and verify results.