physics-informed

This paper proposes a domain-informed adaptive sampling framework for PINNs in metal additive manufacturing, using a conditional Flow Matching model to target high-residual collocation regions, achieving a 62.1% average reduction in relative L2 error over state-of-the-art baselines under an identical collocation budget.

The paper introduces ORACLE, an optimizer-relative constrained learning method that assesses constraint compatibility on the post-optimizer update rather than the objective or gradient, showing improvements over native optimizers in 94% of configurations across eight PDE benchmarks.

This paper proposes a physics-informed world model that generates dynamics from the gradient of a learned energy via a fixed reversible operator, with one-way energy dissipation and structured parameter interventions. Across electromagnetic, particle-in-cell, and shallow-water systems, ~9k-parameter models achieve stable 100x-horizon rollouts, correct counterfactual responses, and out-of-distribution refusal when no intervention map is supplied, far outperforming unstructured equal-capacity baselines.

This arXiv paper identifies 'geometry–physics confounding' as a unified failure mechanism for learning PDEs across varying domains, and proposes a de-confounding framework using geometry-induced coefficient fields and geometry-complete candidate libraries that dramatically improves operator learning and equation discovery accuracy.

A single-authored arXiv paper proposes a data-free physics-informed neural operator for level-set interface advection, trained only on transport residuals and a geometric constraint without any reference solutions, outperforming supervised baselines in area conservation while quantifying the accuracy cost of refusing labels.

KernelOnet proposes an interpretable neural operator framework that explicitly incorporates kernel functions in place of trunk networks, offering data-driven, physics-informed, and hybrid kernel strategies for PDE solving with higher accuracy and fewer parameters than DeepONet.

The paper introduces the Transformer-Mamba for Flow Field (TM4FF) framework, a physics-constrained operator learning model that enhances accuracy and robustness in computational fluid dynamics simulations through innovations like Residual Wavelet Mamba and physics-informed loss.

The paper proposes a Physics Informed Recurrent Neural Network (PIRNN) that predicts unobservable physical variables to improve time series forecasting in physical processes, demonstrated through groundwater level predictions.

Drift Field Net (DFN) is a deep neural network that predicts ocean surface flow fields from satellite observations, using a two-stage training strategy to improve particle trajectory accuracy. It reduces positioning errors by 20 km in 7-day forecasts compared to operational models, with further gains from Lagrangian fine-tuning.

The paper introduces Physics-Informed Conformal Prediction (PI-CP), a framework that embeds PDE residuals into conformal prediction to provide distribution-free, spatially adaptive uncertainty quantification for neural operators like the Fourier Neural Operator (FNO).

The paper introduces Fundamental Dynamical Units (FDUs) as composable primitives for physics-informed structural inference in networked dynamical systems, using neural ODEs to recover interaction structures from perturbation time-series data.

GenONet is a novel deep learning architecture that uses Deep Operator Networks within a Generative Adversarial Network framework for high-resolution precipitation nowcasting, producing sharp and physically consistent forecasts.

The paper proposes a self-augmented diffusion guidance method to incorporate physical laws into diffusion models, reducing deviations from true dynamics and enabling faster generation.

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.

A novel AI model using a Fourier neural operator variant enables density functional theory to scale nearly linearly with system size, allowing efficient simulations of large quantum systems like a magnesium dislocation with 80k electrons on a single GPU.

PhysCaP is a physics-informed code-generation agent that actively explores objects to infer hidden physical properties for efficient robotic manipulation.

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.

PhysAttNet is a physics-informed attention framework that augments lightweight CNN forecasters with domain-guided regularization to improve accuracy and generalization in industrial and astrophysical time series forecasting.

This preprint proposes PI-HNO, a physics-informed hybrid neural operator for transient magnetization prediction in power magnetics, achieving low B-H energy consistency errors with only 4,777 trainable parameters per material model.

This paper evaluates Kolmogorov–Arnold Networks (KANs) versus MLPs as residual branches in hard-constrained recurrent physics-informed networks on an embedded RISC-V platform, finding KANs run slower, consume more energy, and are less dependable under INT8 quantization.