Autoregressive PDE foundation models enable fast full-field forecasting but accumulate errors as predicted states are recursively reused. We introduce Event-Triggered Context Healing (ETCH), which compares forecasts with measurements from 12 cylinder-surface pressure taps. When their discrepancy exceeds a threshold, pressure-conditioned hourglass diffusion transformers reconstruct the velocity--pressure context for subsequent predictions. On 43 RealPDEBench Cylinder trajectories, ETCH reduces velocity and pressure errors from 48.05 and 63.71% to 6.29 and 7.25% for DPOT, and from 74.97 and 76.91% to 5.16 and 6.84% for Poseidon-T, using the same reconstructors and threshold. For DPOT, correcting 14.56% of windows maintains errors close to reconstructing every window and supports average real-time forecasting and correction on an A100 GPU. These results show that sparse physical feedback stabilizes long autoregressive rollouts across PDE foundation models while reducing reconstruction cost.

Real-time monitoring of safety-critical interior states remains an open problem across energy, environmental, and industrial systems where direct instrumentation is infeasible. Existing approaches based on governing equations, discrete state vectors, or fixed sensor locations fail to provide mesh-independent, field-level reconstructions at arbitrary interior coordinates under real-time constraints. Here we introduce neural operator-based virtual sensing as a general framework for recovering inaccessible interior fields from sparse, heterogeneous measurements and operating conditions, instantiated with MIMONet, a multi-input, multi-output neural operator that fuses heterogeneous inputs and decodes coupled physical fields through a shared latent representation. Across three engineering-grade evaluations of escalating complexity, from confined recirculating flows to pressurized water reactor subchannels and compact power-system heat exchangers, MIMONet achieves ≤5% relative reconstruction error, while generating full-field predictions with sub-millisecond inference on an NVIDIA H200, at least five orders of magnitude faster than traditional high-fidelity CFD simulations, with calibrated uncertainty and noise resilience demonstrated on the heat-exchanger case. We validate the framework on three independent real-world datasets spanning electrochemical energy, atmospheric science, and physical oceanography: current-density mapping in a hydrogen fuel cell, wind-speed profiling on instrumented meteorological towers, and North Atlantic ocean-state from a global ocean reanalysis. MIMONet improves over classical virtual-sensing baselines by 72–84%: 84% on fuel-cell internal-field recovery from 10% sensor coverage and 83% on hub-height wind-speed recovery from lower-level tower sensors. Most notably, it recovers sea-surface height, a field with zero direct sensors, from its learned coupling to observed temperature and salinity at 4.1 × lower error than the best classical predictor, a cross-modal field recovery beyond the reach of classical interpolation. These results establish operator-based virtual sensing as a practical route to real-time field observability in systems where the states that matter most cannot be directly measured.

Deep neural operators learn mappings between input functions and complete PDE solution fields, enabling forward evaluations of new problem instances orders of magnitude faster than conventional numerical solvers. Attention mechanisms have recently been introduced into neural operators, but most studies change several architectural components at once, making it difficult to identify what actually improves accuracy. This work presents a controlled and systematic study of five deep operator network (DeepONet) variants with distinct attention mechanisms, trained under both data-driven and physics-informed regimes, to isolate the effects of cross-attention, self-attention, tokenization, and attention depth. We evaluate them on a source-driven transient one-dimensional nonlinear diffusion-reaction equation, a transient one-dimensional viscous Burgers equation with variable initial conditions, and a two-dimensional Poisson heat-conduction problem with heterogeneous source fields. Per-sensor tokenization with cross-attention reduces the mean relative L_2 error of the classical DeepONet in all benchmark-training combinations by factors of 2.4-28.0, while the best attention configurations reach 3.5-32.3. Branch self-attention paired only with dot-product fusion is inconsistent, degrading the one-dimensional problems while helping the more complex two-dimensional source field; added on top of cross-attention it improves all six cases, though by less than cross-attention fusion alone. Global pre-mixing provides no consistent benefit. Increasing cross-attention depth further improves accuracy, but with diminishing returns and a substantially higher cost under physics-informed training. Overall, query-dependent cross-attention is the most reliable mechanism, whereas branch self-attention is most useful for large, spatially complex functional inputs.

Real-time prediction of complex physical systems requires surrogate models that learn from data while representing strong multiphysics coupling. Deep Operator Networks have shown success in single-physics problems, yet their effectiveness in capturing nonlinear interactions in coupled systems (such as thermo-mechanical or electro-thermal coupling) remains underexplored. Here we pose a practical question: should the architecture of a neural operator reflect the strength of physical coupling it aims to model? We compare single-branch and multi-branch designs, in both feedforward and sequential recurrent forms, across three representative systems: a reaction-diffusion problem with heterogeneous sources, a nonlinear thermo-electrical problem with temperature-dependent conductivity and Joule heating, and a viscoplastic thermo-mechanical model of steel solidification. Single-branch networks consistently outperform multi-branch variants in tightly coupled regimes by encouraging shared latent representations, whereas multi-branch designs remain favorable for decoupled or single-physics tasks. Once trained, these surrogates deliver full-field predictions up to 1.8 × 104 times faster than physics-based solvers.

The prevailing paradigm in AI for physical systems: scaling general-purpose foundation models toward universal multimodal reasoning, confronts a barrier at the control interface. Frontier vision–language models achieve only 50–53% accuracy on basic quantitative physics tasks, behaving as approximate guessers that preserve semantic plausibility while violating physical constraints. Safety-critical control demands outcome-space guarantees over executed actions, not parameter-space imitation. Here, we present what is, to our knowledge, the first demonstration of Agentic Physical AI for nuclear reactor control: a domain-specific foundation model in which a compact language model learns control policies through physics-based simulator validation rather than perceptual inference. The framework unifies three capabilities: Agentic AI generates and selects among admissible control strategies; Physical AI evaluates those strategies through closed-loop execution in a reactor simulator; and a domain-specific foundation model acquires reusable control priors through data scaling. Under nominal simulated conditions, scaling yields regime-dependent gains in reliability, including an approximately 500-fold reduction in outcome variance and elimination of terminal-power excursions above 10% on the sampled distribution. Although trained with balanced exposure to four actuation families, the model concentrates 95% of runtime execution on a single-bank strategy without reinforcement learning or reward engineering. The learning framework transfers across reactor simulators without architectural redesign. These results establish a pathway toward reusable, physics-validated foundation-model intelligence for safety-critical control.

Neural operators have emerged as powerful tools for learning nonlinear mappings between function spaces, enabling real-time prediction of complex dynamics in diverse scientific and engineering applications. With their growing adoption in engineering design evaluation, a wide range of neural operator architectures have been proposed for various problem settings. However, model selection remains challenging due to the absence of fair and comprehensive comparisons. To address this, we propose and standardize six representative 3D industry-scale engineering design datasets spanning thermal analysis, linear elasticity, elasto-plasticity, time-dependent plastic problems, and computational fluid dynamics. All datasets include fully preprocessed inputs and outputs for model training, making them directly usable across diverse neural operator architectures. Using these datasets, we conduct a systematic comparison of four types of neural operator variants, including Branch-Trunk-based Neural Operators inspired by DeepONet, Graph-based Neural Operators inspired by Graph Neural Networks, Grid-based Neural Operators inspired by Fourier Neural Operators, and Point-based Neural Operators inspired by PointNet. We further introduce practical enhancements to adapt these models to different engineering settings, improving the fairness of the comparison. Our benchmarking study evaluates each model strengths and limitations in terms of predictive performance, computational efficiency, memory usage, and deployment complexity. The findings provide actionable insights to guide future neural operator development.

Video-diffusion models have recently set the standard in video generation, inpainting, and domain translation thanks to their training stability and high perceptual fidelity. Building on these strengths, we repurpose conditional video diffusion as a physics surrogate for spatio-temporal fields governed by partial differential equations (PDEs). Our two-stage surrogate first applies a Sequential Deep Operator Network (S-DeepONet) to produce a coarse, physics-consistent prior from the prescribed boundary or loading conditions. The prior is then passed to a conditional video diffusion model that learns only the residual: the point-wise difference between the ground truth and the S-DeepONet prediction. By shifting the learning burden from the full solution to its much smaller residual space, diffusion can focus on sharpening high-frequency structures without sacrificing global coherence. The framework is assessed on two disparate benchmarks: (i) vortex-dominated lid-driven cavity flow and (ii) tensile plastic deformation of dogbone specimens. Across these data sets the hybrid surrogate consistently outperforms its single-stage counterpart, cutting the mean relative L2 error from 4.57% to 0.83% for the flow problem and from 4.42% to 2.94% for plasticity, a relative improvements of 81.8% and 33.5% respectively. The hybrid approach not only lowers quantitative errors but also improves visual quality, visibly recovering fine spatial details. These results show that (i) conditioning diffusion on a physics-aware prior enables faithful reconstruction of localized features, (ii) residual learning reduces the problem, accelerating convergence and enhancing accuracy, and (iii) the same architecture transfers seamlessly from incompressible flow to nonlinear elasto-plasticity without problem-specific architectural modifications, highlighting its broad applicability to nonlinear, time-dependent continua.

Scientific applications increasingly demand real-time surrogate models that can capture the behavior of strongly coupled multiphysics systems driven by multiple input functions, such as in thermo-mechanical and electro-thermal processes. While neural operator frameworks, such as Deep Operator Networks (DeepONets), have shown considerable success in single-physics settings, their extension to multiphysics problems remains poorly understood. In particular, the challenge of learning nonlinear interactions between tightly coupled physical fields has received little systematic attention. This study addresses a foundational question: should the architectural design of a neural operator reflect the strength of physical coupling it aims to model? To answer this, we present the first comprehensive, architecture-aware evaluation of DeepONet variants across three regimes: single-physics, weakly coupled, and strongly coupled multiphysics systems. We consider a reaction-diffusion equation with dual spatial inputs, a nonlinear thermo-electrical problem with bidirectional coupling through temperature-dependent conductivity, and a viscoplastic thermo-mechanical model of steel solidification governed by transient phase-driven interactions. Two operator-learning frameworks, the classical DeepONet and its sequential GRU-based extension, S-DeepONet, are benchmarked using both single-branch and multi-branch (MIONet-style) architectures. Our results demonstrate that architectural alignment with physical coupling is crucial: single-branch networks significantly outperform multi-branch counterparts in strongly coupled settings, whereas multi-branch encodings offer advantages for decoupled or single-physics problems. Once trained, these surrogates achieve full-field predictions up to 1.8e4 times faster than high-fidelity finite-element solvers, without compromising solution accuracy.

`Real-time prediction of complex physical systems requires surrogate models that learn from data while representing strong multiphysics coupling. Deep Operator Networks have shown success in single-physics problems, yet their effectiveness in capturing nonlinear interactions in coupled systems (such as thermo-mechanical or electro-thermal coupling) remains underexplored. Here we pose a practical question: should the architecture of a neural operator reflect the strength of physical coupling it aims to model? We compare single-branch and multi-branch designs, in both feedforward and sequential recurrent forms, across three representative systems: a reaction--diffusion problem with heterogeneous sources, a nonlinear thermo-electrical problem with temperature-dependent conductivity and Joule heating, and a viscoplastic thermo-mechanical model of steel solidification. Single-branch networks consistently outperform multi-branch variants in tightly coupled regimes by encouraging shared latent representations, whereas multi-branch designs remain favorable for decoupled or single-physics tasks. Once trained, these surrogates deliver full-field predictions up to $1.8 \times 10^4$ times faster than physics-based solvers.

Partial differential equations (PDEs) are fundamental to modeling complex and nonlinear physical phenomena, but their numerical solution often requires significant computational resources, particularly when a large number of forward full solution evaluations are necessary, such as in design, optimization, sensitivity analysis, and uncertainty quantification. Recent progress in operator learning has enabled surrogate models that efficiently predict full PDE solution fields; however, these models often struggle with accuracy and robustness when faced with highly nonlinear responses driven by sequential input functions. To address these challenges, we propose the Sequential Neural Operator Transformer (S-NOT), a architecture that combines gated recurrent units (GRUs) with the self-attention mechanism of transformers to address time-dependent,nonlinear PDEs. Unlike S-DeepONet (S-DON), which uses a dot product to merge encoded outputs from the branch and trunk sub-networks, S-NOT leverages attention to better capture intricate dependencies between sequential inputs and spatial query points. We benchmark S-NOT on three challenging datasets from real-world applications with plastic and thermo-viscoplastic highly nonlinear material responses: multiphysics steel solidification, a 3D lug specimen, and a dogbone specimen under temporal and path-dependent loadings. The results show that S-NOT consistently achieves a higher prediction accuracy than S-DON even for data outliers, demonstrating its accuracy and robustness for drastically accelerating computational frameworks in scientific and engineering applications.

Accurate reconstruction of latent environmental fields from sparse, indirect observations is a fundamental challenge across scientific domains, from atmospheric science and geophysics to public health and aerospace safety. Existing approaches typically rely on physics-based simulations or dense sensor networks; however, these methods are hampered by high computational cost, latency, and limited spatial coverage. Here we introduce the \textbf{Temporal Radiation Operator Network (TRON)}, a spatiotemporal neural operator architecture that infers continuous global scalar fields solely from sequences of sparse, non-uniform proxy measurements. Unlike recent prediction models that require dense, gridded inputs to predict system states, TRON tackles sparse-to-dense, cross-domain field reconstruction. It reconstructs the current global field in real time from sparse, temporally evolving sensor data, without access to any future observations or dense ground-truth fields. We demonstrate this approach on global cosmic radiation dose mapping: TRON, trained on daily reference fields spanning 2001 to 2023, generalizes across 65,341 spatial locations with input sequences ranging from 7 to 90 days. It achieves sub-second inference with relative $L_2$ errors below 0.1\%, representing over a 58,000$\times$ speedup compared to physics-based estimators. Although showcased in the radiation application, TRON provides a domain-agnostic framework for continuous field reconstruction from sparse data, with broad applications in atmospheric modeling, geophysical hazard monitoring, and real-time environmental risk prediction.