betaFeedback

WorksComputer Science

AutoPDEBench: Benchmarking LLM Auto-Research for Neural PDE Solver Design

Abstract

Partial differential equations (PDEs) are essential for modeling complex physical systems, and neural solvers have recently emerged as powerful data-driven tools for numerically solving them. However, existing neural solvers struggle with domain-specific challenges, such as varying parameters and high-speed flows, necessitating specialized architectures. Manually designing these specialized solver architectures is a highly iterative, time-consuming process requiring deep expertise, creating a significant bottleneck in scientific discovery. We propose leveraging autonomous AI research agents to automate the synthesis of specialized solvers. To support this, we introduce AutoPDEBench, a benchmark dedicated to LLM-driven automated research for PDE solver design. The benchmark includes 25 challenging datasets featuring both novel and actively studied physical scenarios. We evaluate a suite of general-purpose models (transformer, ROM, and graph-based) alongside a multi-agent instantiation of the iterative automated research pipeline, which serves as an agentic baseline. Empirical results show that the iterative automated research system significantly outperforms the general-purpose neural solver baselines. Our findings demonstrate the viability of using AI agents to automatically design neural solvers for complex physical systems. AutoPDEBench provides a foundational testbed to accelerate agent-driven scientific discovery in physics and engineering.

Introduction

Partial differential equations (PDEs) are the foundational language of physics. They provide the essential mathematical framework to describe how complex systems evolve over space and time. Recent advancements have seen neural solvers demonstrate remarkable performance in learning PDE solutions purely from data. While these models achieve strong results across many standard datasets, others present domain-specific challenges that necessitate specialized architectures. For example, when evaluating datasets with mixed PDE parameters, standard neural solvers cannot reliably map initial conditions to target trajectories; the lack of explicit parameter inputs renders the predictions ill-posed. Encountering this exact scenario, Takamoto et al. (2023) recognized that solvers must explicitly encode the underlying PDE parameters. To address this, they introduced a parameter-guided channel attention mechanism that conditions the neural solver on these varying physical coefficients. High-speed flows present another domain-specific challenge. Because these systems involve extreme gradients and discontinuities, such as shock waves, existing neural solvers typically suffer from numerical instability or non-physical smoothing. To overcome this, Helwig et al. [26] introduced NeuralCFL, a framework that explicitly predicts an adaptive timestep size to stabilize the simulation during rapid state changes.

Traditionally, human researchers invest a substantial amount of time designing these domain-specific solvers, a process that demands deep specialized expertise and significant computational resources. Because this manual engineering is highly iterative and uniquely tailored to each physical system, it severely limits the pace and scalability of scientific discovery. This critical bottleneck motivates the development of autonomous AI research agents capable of automatically synthesizing specialized solver pipelines. By dynamically adapting to complex physical dynamics, these agents promise to accelerate the modeling process and reduce the reliance on human intervention.

To this end, we introduce AutoPDEBench, a benchmark dedicated to LLM-driven automated research for designing specialized solvers to tackle domain-specific challenges. AutoPDEBench comprises 25 challenging datasets, encompassing both actively studied and entirely novel physical scenarios. These datasets present distinct, idiosyncratic challenges that general-purpose neural solvers either struggle with or fail to handle entirely, thereby necessitating specialized architectures. To establish robust baselines, we first evaluate a suite of versatile general-purpose models, including transformer-based, reduced-order modeling (ROM), and graph-based methods. We then introduce a multi-agent instantiation of the iterative automated research pipeline to serve as an agentic baseline. Our empirical results demonstrate that this auto-research system achieves superior performance compared to the general solver baselines, underscoring the immense potential of utilizing AI agents to automatically design neural solvers for complex physical systems. However, outperforming general-purpose baselines does not mean the task is fully solved. In classical numerical analysis, the search for schemes with minimal local truncation error is an ongoing pursuit, as per-step errors inevitably accumulate over long time horizons and are often amplified by nonlinear dynamics. Neural solvers face this exact vulnerability, with single-step prediction errors compounding severely during autoregressive rollouts. We therefore position AutoPDEBench as an open-ended testbed designed to continually challenge agents to discover increasingly accurate solvers.

Our contributions are as follows:

  • Problem Identification: We highlight that the domain of neural PDE solvers encompasses a diverse array of physical systems characterized by idiosyncratic challenges. Traditionally, addressing these unique constraints has demanded extensive manual effort to design specialized architectures.

  • Proposed Benchmark: We introduce AutoPDEBench, the first comprehensive benchmark featuring 25 challenging datasets with unique physical constraints. This collection spans both actively researched problems and entirely novel domains introduced to the community for the first time.

  • Empirical Analysis: We conduct extensive evaluations comparing versatile, general-purpose neural PDE solvers against a multi-agent instantiation of the iterative automated research pipeline. Our results demonstrate the profound potential of utilizing LLM agents to autonomously design specialized solver pipelines for complex physical systems.

Related Work

We identify three key areas closely related to our work.

Neural PDE Solvers. The evolution of neural PDE solvers originated with early architectures designed to learn solution operators on regular Euclidean grids [39, 33] (Lu et al. 2021; Li et al. 2021; Rahman et al. 2023; Tripura & Chakraborty 2023; Tran et al. 2023). To overcome the geometric limitations of uniform grid-based representations, subsequent research rapidly expanded model capabilities to accommodate unstructured meshes [4, 22] (Pfaff et al. 2021; Li et al. 2023; Lino et al. 2021) and Lagrangian particle-based simulations (Sanchez-Gonzalez et al., 2020; Toshev et al., 2024b; 2023; Prantl et al., 2022; Rochman-Sharabi et al., 2025). As these methodologies matured, the community introduced comprehensive benchmark suites [21, 32] (Wang et al. 2026a; Toshev et al. 2024a; Ohana et al. 2024; Takamoto et al. 2024) to establish standardized evaluation strategies and curate large-scale datasets across a wide variety of physical systems. Building upon these established baselines, recent solvers have sought to push performance boundaries across multiple complex axes. These advancements include architectures capable of generalizing across varying PDE parameters [5] (Takamoto et al. 2023) and diverse boundary conditions [28] (Wang et al. 2024; Saad et al. 2023), handling highly complex geometries (Wu et al., 2024; Luo et al.; Wen et al., 2025; Mousavi et al., 2025), adapting seamlessly across 1D, 2D, and 3D spatial dimensions [9], and unifying Eulerian field and Lagrangian particle representations [2, 3]. Most recently, this trajectory has culminated in the exploration of physics foundation models [25, 27] (McCabe et al. 2024; Ye et al. 2024; Shen et al. 2024; Zhou et al.; Wang et al. 2025), which are pre-trained on massive, heterogeneous datasets to achieve strong zero-shot and few-shot generalization capabilities across unseen dynamic regimes.

Beyond forward temporal simulation, neural network methodologies have been widely adopted across adjacent domains in scientific machine learning. These broader applications include solving inverse problems [36, 35] (Zheng et al. 2025; Molinaro et al. 2023; Wang & Wang 2024) and developing hybrid frameworks that integrate neural networks directly into classical numerical schemes to enhance their performance and mitigate truncation errors [19] (List et al. 2022; Sun et al. 2023; Sappl et al. 2019).

LLM for PDE Solvers. Researchers have extensively utilized LLMs to autonomously generate traditional numerical PDE solvers [37, 23, 12] (Wu et al. 2025). These approaches typically translate high-level mathematical formulations and boundary conditions directly into executable simulation scripts. This automation significantly reduces the manual programming effort and tedious debugging traditionally required for complex scientific computing. Recent studies have extended this paradigm to neural architectures. For example, Wuwu et al. (2025) leverage LLM agents to design neural solvers; however, their evaluations remain confined to standard benchmark datasets. Our work diverges significantly by deploying LLM agents to tackle complex datasets characterized by unique physical constraints that require domain-specific solvers. Because our benchmark introduces entirely novel, previously unstudied physical systems, the LLM cannot simply retrieve memorized templates from its pre-training data. Instead, the agent must autonomously identify the underlying physical bottlenecks and synthesize novel architectural solutions to overcome them.

LLM Auto Research. The integration of Large Language Models (LLMs) into the scientific workflow has evolved along a spectrum of increasing autonomy. Initially, research focused on utilizing LLMs as assistive tools to augment human effort in specific, localized tasks, such as conducting literature reviews [1, 14, 18] (Skarlinski et al. 2024), generating preliminary ideas (Yang et al., 2025; Qiu et al., 2025; Ghaffarollahi & Buehler, 2024), and assisting with experimental planning and iteration [13, 20]. As their capabilities expanded, LLMs transitioned into more autonomous roles, functioning as analytical agents capable of complex information processing with minimal human intervention. In this capacity, they have been deployed for automated machine learning task modeling [30, 34, 7] and the identification of governing equations from observational data [10] (Merler et al. 2024). Most recently, the field has advanced toward the LLM scientists paradigm. These systems operate as general-purpose, active agents capable of orchestrating the entire scientific discovery process end-to-end [6, 16] (Mitchener et al. 2025; Novikov et al. 2025; Tang et al. 2025; Schmidgall et al. 2025). By independently navigating from novel ideation and autonomous experimental execution to synthesizing genuine findings into full research paper drafts, these pipelines represent a significant step toward fully autonomous natural science discovery.

AutoPDEBench

Problem Formulation

We present the overall framework in Figure 2. We consider the simulation of a physical system where our goal is to learn a neural simulator Φθ\Phi_\theta that predicts a future output state y∈Yy \in\mathcal{Y} from an input state x∈Xx \in\mathcal{X}. The model also accepts optional supplementary information c∈Cc \in\mathcal{C}, which can include a variable timestep Δt\Delta t, boundary conditions, PDE parameters, or the governing PDE equations. Thus, the simulator is defined as Φθ:X×C→Y\Phi_\theta: \mathcal{X} \times\mathcal{C} \to\mathcal{Y}, yielding the prediction rule:

Overall framework of AutoPDEBench

Figure 2. Overall framework of AutoPDEBench. We propose 25 challenging datasets, each presenting unique domain-specific challenges that necessitate specialized solvers. While general-purpose neural solvers can often be applied, they inherently lack the mechanisms to handle these specific constraints, leading to suboptimal performance. Traditionally, researchers manually design specialized solvers for these tasks, which requires extensive domain expertise and is highly time-consuming. We argue that automated research agents offer a compelling alternative, leveraging LLMs to autonomously design neural solvers capable of addressing these dataset-specific challenges.

y=Φθ(x,c).(1)y = \Phi_\theta(x,c). \tag*{(1)}

Each dataset is split into disjoint Dtrain\mathcal{D}_{\mathrm{train}}, Dval\mathcal{D}_{\mathrm{val}}, and Dtest\mathcal{D}_{\mathrm{test}}, and L(Φθ;D)\mathcal{L}(\Phi_\theta;\mathcal{D}) denotes evaluation metrics evaluated on split D\mathcal{D}. We evaluate models under a strict hardware constraint. Given a fixed GPU with memory budget BB, our objective is to achieve the best possible performance. We place no restrictions on training or inference time, FLOPs, or parameter count.

General-purpose solver. Suppose h∗h^* is the optimal hyperparameter configuration evaluated on the validation dataset that respects the memory budget BB:

h∗=arg⁡min⁡h:Mem⁡(Φ,h)≤B L(Φθh;Dval),θh∗=Train⁡(Φ,h∗,Dtrain).(2)h^* = \underset{h:\operatorname{Mem}(\Phi,h)\le B}{\arg\min}\ \mathcal{L}(\Phi_{\theta_h};\mathcal{D}_{\mathrm{val}}), \qquad\theta_{h^*} = \operatorname{Train}(\Phi,h^*,\mathcal{D}_{\mathrm{train}}). \tag*{(2)}

We then use this configuration to report the final test performance L(Φθh∗;Dtest)\mathcal{L}(\Phi_{\theta_{h^*}};\mathcal{D}_{\mathrm{test}}).

Auto-research agents. We abstract the research process as an agentic system A\mathcal{A} that, at each round, maps a context I\mathcal{I} to a candidate solution. The context I\mathcal{I} aggregates the information available to the agent, such as the dataset description and the modeling objective. The agentic system is a mapping A:I→M×F×H\mathcal{A}:\mathcal{I}\to\mathcal{M}\times\mathcal{F}\times\mathcal{H} producing a model file m∈Mm\in\mathcal{M} (architecture Φm\Phi_m), a loss file ℓ∈F\ell\in\mathcal{F} (training objective Ltrain\mathcal{L}^{\mathrm{train}}, which governs optimization but does not alter the fixed evaluation metrics), and a config file h∈Hh\in\mathcal{H} (hyperparameters). Over rounds i=1,…,Ni=1,\ldots,N, the context Ii\mathcal{I}_i is updated with the latest history, and the agent produces

(mi,ℓi,hi)=A(Ii),θi=Train⁡(Φmi,Ltrain,hi,Dtrain),(3)(m_i,\ell_i,h_i)=\mathcal{A}(\mathcal{I}_i),\qquad\theta_i=\operatorname{Train}(\Phi_{m_i},\mathcal{L}^{\mathrm{train}},h_i,\mathcal{D}_{\mathrm{train}}), \tag*{(3)}

subject to the memory limit Mem⁡(Φmi,hi)≤B\operatorname{Mem}(\Phi_{m_i},h_i)\le B. Each candidate is trained on Dtrain\mathcal{D}_{\mathrm{train}} under its generated objective Ltrain\mathcal{L}^{\mathrm{train}} and hyperparameters hih_i, and the agent observes only the validation outcome ri=L(Φmi,θi;Dval)r_i=\mathcal{L}(\Phi_{m_i,\theta_i};\mathcal{D}_{\mathrm{val}}), never any quantity on Dtest\mathcal{D}_{\mathrm{test}}. After NN rounds, it commits its best validated configuration i∗=arg⁡min⁡irii^*=\arg\min_i r_i, and the reported performance is its evaluation on the held-out test split, L(Φmi∗,θi∗;Dtest)\mathcal{L}(\Phi_{m_{i^*},\theta_{i^*}};\mathcal{D}_{\mathrm{test}}).

Datasets

We propose 25 challenging datasets, each presenting unique constraints that necessitate domain-specific solvers. This collection includes both entirely novel and currently understudied physical systems. Example visualizations and overarching descriptions are provided in Figure 1 and Table 1, respectively. For comprehensive details regarding data generation, governing physics, problem setups, and additional visualizations, we refer readers to the dedicated appendices for each dataset.

Overview of the example datasets proposed by AutoPDEBench

Figure 1. Overview of the example datasets proposed by AutoPDEBench. AutoPDEBench consists of 25 challenging datasets, each presenting specific domain challenges and necessitating specialized solvers. It features both entirely novel and currently understudied datasets. We intentionally reduced the resolution of all figures to maintain a reasonable file size. High-quality versions are available in our GitHub repository.

DatasetDescriptionSource
HighSpeedFlow §I.1Supersonic flows that exceed the speed of sound and create sudden changes such as shock waves.Public
RealData §I.2Real-world data consists of noisy measurements collected using time-resolved PIV.Public
MixedParam §I.3Datasets with mixed PDE parameters, where the governing coefficients vary across trajectories.Ours
Heterogeneous §I.4Lagrangian particle datasets with particles possessing distinct physical attributes such as density.Ours
Partial §I.5Datasets with partially observed initial conditions that require predicting the full trajectory.Ours
MixedDomain §I.6Datasets featuring grid, mesh, and particle representations of the same underlying physical system.Ours
MultiPhysics §I.7Multiphysics simulations that capture the interactions between multiple physical processes.Public
3D §I.8Spatiotemporal datasets featuring three-dimensional volumetric observations.Public
IrregularTime §I.9Datasets featuring observations recorded at non-uniform time intervals.Ours
LagInletOutlet §I.10Lagrangian inlet-outlet data lacking 1-to-1 particle correspondence.Ours
LagFSI §I.11Lagrangian fluid-solid interaction datasets where both phases are represented by Lagrangian particles.Ours
Extreme §I.12Datasets with extreme scale capturing 3D supernova blastwaves across seven orders of magnitude.Public
MixedRes §I.13Generalization datasets that capture PDE dynamics on diverse domains and resolutions.Ours
ExtLag1 §I.14Lagrangian particle datasets with liquid sloshing as external force.Ours
ExtLag2 §I.15Datasets capturing Lagrangian fluid dynamics driven by the external force of a moving piston.Ours
MixedDim §I.16Datasets capturing identical PDE dynamics in 1D, 2D, and 3D.Ours
MixedBC §I.17Datasets capturing identical PDE under Dirichlet, Neumann, and periodic boundary conditions.Ours
LongHorizon §I.18Datasets featuring extremely long-horizon, 5,000-step rollouts.Ours
Stochastic §I.19Stochastic PDE datasetOurs
MultiPhase §I.20Lagrangian particle datasets modeling multiphase liquid-gas interactions.Ours
MixedPDE §I.21Datasets comprising a diverse range of PDEs,Ours
LargeLag §I.22Large-scale, high-resolution Lagrangian dataset featuring millions of particles.Ours
Sim2Real §I.23Datasets designed for sim-to-real transfer.Public
HighDimensional §I.245D gyrokinetic dataset capturing high-dimensional PDE dynamics.Ours
ComplexGeometry §I.25Datasets featuring flow around complex geometries.Public

Table 1. Overview of datasets in AutoPDEBench. Detailed descriptions and visualizations for each dataset are provided in their individual Appendix section.

Evaluation Protocol

Our evaluation protocol assesses final model performance on the test dataset. Given our GPU resource constraints, each model was tuned to maximize performance within these limits. Our primary evaluation metric is RMSE. We also use dataset-specific metrics to evaluate additional aspects such as physical consistency and practical utility. The additional evaluation results for each dataset are provided in their individual appendix sections, while detailed descriptions of these supplementary metrics can be found in Appendix C.

General-Purpose Neural Solver Baselines

To establish our baselines, we consider transformer-based, reduced-order modeling, and graph-based neural solvers. We selected these architectural paradigms for their broad generality and structural flexibility. Unlike convolution-based methods, which are fundamentally constrained by the requirement for uniform spatial grids, these selected models feature adaptable input-output mappings. This flexibility allows them to universally process highly complex, unstructured, and dimensionally diverse data representations. Given their capacity to handle such generalized inputs, we select GNOT [24], Transolver (Wu et al., 2024), Transolver++ (Luo et al.), LNO (Wang & Wang, 2024), and AMG [38] as our representative baselines.

Foundation models, pretrained on massive amounts of data, would theoretically serve as ideal baselines. However, current foundation models [25, 27] (McCabe et al. 2024; Ye et al. 2024; Shen et al. 2024; Zhou et al.; Wang et al. 2025) are predominantly restricted to grid-based domains. They lack exposure to the Eulerian mesh and Lagrangian particle representations that are central to our work. Furthermore, their rigid input-output architectures preclude them from generalizing across our highly unstructured data, which features diverse spatial dimensions and varied discretization schemes.

Multi-Agent Instantiation of Auto Research Pipeline

AutoPDEBench is agnostic to the internal design of A\mathcal{A}. Any system that maps Ii\mathcal{I}_i to (mi,ℓi,hi)(m_i,\ell_i,h_i) under the protocol above can be evaluated, including more elaborate multi-agent frameworks with dynamic delegation or parallel search. We provide a reference instantiation of A\mathcal{A} as a role-specialized multi-agent system of three LLM-driven components: a planner, a coder, and a debugger. The planner consumes the dataset description τdata\tau_{\mathrm{data}}, the modeling objective τobj\tau_{\mathrm{obj}}, the hardware constraints τconst\tau_{\mathrm{const}}, the coding interface specification τinterface\tau_{\mathrm{interface}} (editable files and program structure), and the trial history τhistory\tau_{\mathrm{history}}, and produces a detailed step-by-step plan

τplan=Aplanner(τdata,τobj,τconst,τinterface,τhistory),(4)\tau_{\mathrm{plan}} = \mathcal{A}_{\mathrm{planner}}\left(\tau_{\mathrm{data}}, \tau_{\mathrm{obj}}, \tau_{\mathrm{const}}, \tau_{\mathrm{interface}}, \tau_{\mathrm{history}}\right), \tag*{(4)}

a natural-language specification of the model architecture, loss functions, and hyperparameter choices to attempt this round, together with its rationale. The coder then translates this plan into executable files, conditioning on the hardware constraints, the coding interface specification, and the plan:

(mi,ℓi,hi)=Acoder(τconst,τinterface,τplan).(5)(m_i,\ell_i,h_i) = \mathcal{A}_{\mathrm{coder}}\left(\tau_{\mathrm{const}},\tau_{\mathrm{interface}},\tau_{\mathrm{plan}}\right). \tag*{(5)}

We then attempt θi=Train⁡(Φmi,Ltrain,hi,Dtrain)\theta_i = \operatorname{Train}(\Phi_{m_i},\mathcal{L}^{\mathrm{train}},h_i,\mathcal{D}_{\mathrm{train}}). If training fails, raising an execution error or violating the memory budget Mem⁡(Φmi,hi)≤B\operatorname{Mem}(\Phi_{m_i},h_i) \le B, we invoke the debugger, which revises the files given the current code, the hardware constraints, the coding interface specification, the plan, and the observed error ϵ\epsilon:

(mi,ℓi,hi)←Adebugger(mi,ℓi,hi,τconst,τinterface,τplan,ϵ).(6)(m_i,\ell_i,h_i) \leftarrow\mathcal{A}_{\mathrm{debugger}}\left(m_i,\ell_i,h_i,\tau_{\mathrm{const}},\tau_{\mathrm{interface}},\tau_{\mathrm{plan}},\epsilon\right). \tag*{(6)}

This step repeats until training completes successfully or a fixed retry budget is exhausted, in which case the round is recorded as a failure.

Experiments

We present the main results in Table 2, where we use RMSE as the primary evaluation metric and refer the readers to the Appendix of each dataset for additional metrics. Overall, we find that even this baseline multi-agent system achieves significantly better performance than general-purpose neural solvers. This underscores the value of tailoring architectures to address dataset-specific challenges and validates the efficacy of utilizing autonomous research agents for this task.

However, the agent struggles with highly complex datasets that remain open challenges in the literature. For instance, the LagFEM dataset models a sphere dropping into a body of water, with both phases discretized using Lagrangian particles. This scenario necessitates innovative solutions for two-way fluid-solid coupling, alongside methods to strictly enforce the rigid integrity of the sphere, which are non-trivial challenges that existing Lagrangian neural solvers have yet to successfully resolve. Consequently, the agent’s generated design performs significantly worse than the

References

  1. [1]Shubham Agarwal, Gaurav Sahu, Abhay Puri, Issam H. Laradji, Krishnamurthy DJ Divjotham, Jason Stanley, Laurent Charlin, and Christopher Pal. Litllm: A toolkit for scientific literature review, 2025. URL https://arxiv.org/abs/2402.01788.
  2. [2]Benedikt Alkin, Andreas Fürst, Simon Schmid, Lukas Gruber, Markus Holzleitner, and Johannes Brandstetter. Universal physics transformers. arXiv preprint arXiv:2402.12365, 2024.arxiv.org/abs/2402.12365
  3. [3]Benedikt Alkin, Maurits Bleeker, Richard Kurle, Tobias Kronlachner, Reinhard Sonnleitner, Matthias Dorfer, and Johannes Brandstetter. AB-UPT: Scaling neural CFD surrogates for high- fidelity automotive aerodynamics simulations via anchored- branched universal physics transformers. Transactions on Machine Learning Research, 2025. ISSN 2835-8856. URL https://openreview.net/forum?id=nwQ8nitlTZ.arxiv.org/abs/2502.09692
  4. [4]Oussama Boussif, Yoshua Bengio, Loubna Benabbou, and Dan Assouline. MAgnet: Mesh agnostic neural PDE solver. In Alice H. Oh, Alekh Agarwal, Danielle Belgrave, and Kyunghyun Cho (eds.), Advances in Neural Information Processing Systems, 2022. URL https://openreview.net/forum?id=bx2roi8hc8.arxiv.org/abs/2210.05495
  5. [5]Johannes Brandstetter, Daniel E. Worrall, and Max Welling. Message passing neural PDE solvers. In International Conference on Learning Representations, 2022. URL https://openreview.net/forum?id=vSix3HPYKSU.arxiv.org/abs/2202.03376
  6. [6]Garyk Brixi, Matthew G. Durrant, Jerome Ku, Mohsen Naghipourfar, Michael Poli, Gwanggyu Sun, Greg Brockman, Daniel Chang, Alison Fanton, Gabriel A. Gonzalez, Samuel H. King, David B. Li, Aditi T. Merchant, Eric Nguyen, Chiara Ricci-Tam, David W. Romero, Jonathan C. Schmok, Ali Taghibakhshi, Anton Vorontsov, Brandon Yang, Myra Deng, Liv Gorton, Nam Nguyen, Nicholas K. Wang, Michael T. Pearce, Elana Simon, Etowah Adams, Zachary J. Amador, Euan A. Ashley, Stephen A. Baccus, Haoyu Dai, Steven Dillmann, Stefano Ermon, Daniel Guo, Michael H. Herschl, Rajesh Iango, Ken Janik, Amy X. Lu, Reshma Mehta, Mohammad R. K. Mofrad, Madelena Y. Ng, Jaspreet Pannu, Christopher Ré, John St. John, Jeremy Sullivan, Joseph Tey, Ben Viggiano, Kevin Zhu, Greg Zynda, Daniel Balsam, Patrick Collison, Anthony B. Costa, Tina Hernandez-Boussard, Eric Ho, Ming-Yu Liu, Thomas McGrath, Kimberly Powell, Sudarshan Pinglay, Dave P. Burke, Hani Goodarzi, Patrick D. Hsu, and Brian L. Hie. Genome modelling and design across all domains of life with evo 2. Nature, 2026. doi: 10.1038/s41586-026-10176-5. URL https://doi.org/10.1038/s41586-026-10176-5.
  7. [7]Jun Shern Chan, Neil Chowdhury, Oliver Jaffe, James Aung, Dane Sherburn, Evan Mays, Giulio Starace, Kevin Liu, Leon Maksin, Tejal Patwardhan, Aleksander Madry, and Lilian Weng. MLE-bench: Evaluating machine learning agents on machine learning engineering. In The Thirteenth International Conference on Learning Representations, 2025. URL https://openreview.net/forum?id=6s5uXNWGIh.arxiv.org/abs/2410.07095
  8. [8]Angel X. Chang, Thomas Funkhouser, Leonidas Guibas, Pat Hanrahan, Qixing Huang, Zimo Li, Silvio Savarese, Manolis Savva, Shuran Song, Hao Su, Jianxiong Xiao, Li Yi, and Fisher Yu. Shapnet: An information-rich 3d model repository, 2015. URL https://arxiv.org/abs/1512.03012.
  9. [9]Tianyu Chen, Haoyi Zhou, Ying Li, Hao Wang, Chonghan Gao, Rongye Shi, Shanghang Zhang, and Jianxin Li. Omniarch: Building foundation model for scientific computing. In Forty-second International Conference on Machine Learning, 2025. URL https://openreview.net/forum?id=UlprLwWYKp.arxiv.org/abs/2402.16014
  10. [10]Yimeng Chen, Piotr Piekos, Mateusz Ostaszewski, Firas Laakom, and Jürgen Schmidhuber. Phys- gym: Benchmarking LLMs in interactive physics discovery with controlled priors. In The Thirty- ninth Annual Conference on Neural Information Processing Systems Datasets and Benchmarks Track, 2026. URL https://openreview.net/forum?id=w8uII2qAmd.DOI
  11. [11]J.M. Domínguez, G. Fourtakas, C. Altomare, R.B. Canelas, A. Tafuni, O. García-Feal, I. Martínez-Estévez, A. Mokos, R. Vacondio, A.J.C. Crespo, B.D. Rogers, P.K. Stansby, and M. Gómez-Gesteira. Dualsphysics: from fluid dynamics to multiphysics problems. Computational Particle Mechanics, 9(5):867–895, 2022. ISSN 2196-4386. doi: https://doi.org/ 10.1007/s40571-021-00404-2. URL https://www.sciencedirect.com/science/article/pii/S219643862500631X.
  12. [12]Huanshuo Dong, Keyao Zhang, Hong Wang, Zhezheng Hao, Zhiwei Zhuang, Ziyan Liu, Jiacong Wang, Gengyuan Liu, and Xin Jin. Autopde: Reliable agentic pde solving via explicitly represented solver strategies, 2026. URL https://arxiv.org/abs/2606.10752.
  13. [13]Fabio Ferreira, Lucca Wobbe, Arjun Krishnakumar, Frank Hutter, and Arber Zela. Can llms beat classical hyperparameter optimization algorithms? a study on autoresearch, 2026. URL https://arxiv.org/abs/2603.24647.
  14. [14]Muhan Gao, Jash Shah, Weiqi Wang, Kuan-Hao Huang, and Daniel Khashabi. Science hierarchography: Hierarchical organization of science literature, 2025. URL https://arxiv.org/abs/2504.13834.
  15. [16]Ali Essam Ghareeb, Benjamin Chang, Ludovico Mitchener, Angela Yiu, Caralyn J. Szostkiewicz, Jon M. Laurent, Muhammed T. Razzak, Andrew D. White, Michaela M. Hinks, and Samuel G. Rodrigues. Robin: A multi-agent system for automating scientific discovery, 2025. URL https://arxiv.org/abs/2505.13400.
  16. [17]R. A. Gingold and J. J. Monaghan. Smoothed particle hydrodynamics: theory and application to non-spherical stars. Monthly Notices of the Royal Astronomical Society, 181(3):375–389, 12 1977. ISSN 0035-8711. doi: 10.1093/mnras/181.3.375. URL https://doi.org/10.1093/mnras/181.3.375.
  17. [18]Gregory Go, Khang Ly, Anders Sogaard, Amin Reza Tabatabaei, Maarten de Rijke, and Xinyi Chen. Lira: A multi-agent framework for reliable and readable literature review generation. In AAAI Conference on Artificial Intelligence, 2025. URL https://api.semanticscholar.org/CorpusID:281886200.
  18. [19]Daniel Greenfeld, Meirav Galun, Ronen Basri, Irad Yavneh, and Ron Kimmel. Learning to optimize multigrid PDE solvers. In International Conference on Machine Learning, pp. 2415–2423. PMLR, 2019.arxiv.org/abs/1902.10248
  19. [20]Taicheng Guo, Nitesh V. Chawla, Olaf Wiest, and Xiangliang Zhang. Autollmresearch: Training research agents for automating llm experiment configuration - learning from cheap, optimizing expensive, 2026. URL https://arxiv.org/abs/2605.11518.
  20. [21]Jayesh K Gupta and Johannes Brandstetter. Towards multi-spatiotemporal-scale generalized pde modeling. arXiv preprint arXiv:2209.15616, 2022.arxiv.org/abs/2209.15616
  21. [22]Jiaqi Han, Wenbing Huang, Hengbo Ma, Jiachen Li, Joshua B Tenenbaum, and Chuang Gan. Learning physical dynamics with subequivariant graph neural networks. arXiv preprint arXiv:2210.06876, 2022.arxiv.org/abs/2210.06876
  22. [23]Zhen Hang, Yushan Yashengjiang, Junhui Li, Huanshuo Dong, Yang Wei, Zhezheng Hao, Jiangtao Ma, Songlin Bai, Haozhong Kai, Xihang Yue, Gangzong Si, Dongming Jiang, Chao Yao, Zhanhua Hu, Jiangqing Zhang, Pengwei Liu, Yaomin Shen, Xingyu Ren, Lei Liu, Zikang Xu, Han Li, Qingsong Yao, Hande Dong, and Hong Wang. Pdeagent-bench: A multi-metric, multi-library benchmark for pde solver generation, 2026. URL https://arxiv.org/abs/2605.09636.
  23. [24]Zhongkai Hao, Chengyang Ying, Zhengyi Wang, Hang Su, Yinpeng Dong, Songming Liu, Ze Cheng, Jun Zhu, and Jian Song. Gnot: A general neural operator transformer for operator learning. ArXiv, abs/2302.14376, 2023. URL https://api.semanticscholar.org/CorpusID:257232579.arxiv.org/abs/2302.14376
  24. [25]Zhongkai Hao, Chang Su, Songming Liu, Julius Berner, Chengyang Ying, Hang Su, Anima Anandkumar, Jian Song, and Jun Zhu. Dpot: Auto-regressive denoising operator transformer for large-scale pde pre-training. arXiv preprint arXiv:2403.03542, 2024.arxiv.org/abs/2403.03542
  25. [26]Jacob Helwig, Sai Sreeharsha Adavi, Xuan Zhang, Yuchao Lin, Felix S. Chim, Luke Takeshi Vizzini, Haiyang Yu, Muhammad Hasnain, Saykat Kumar Biswas, John J. Holloway, Narendra Singh, N. K. Anand, Swagnik Guhathakurta, and Shuiwang Ji. A two-phase deep learning framework for adaptive time-stepping in high-speed flow modeling, 2026. URL https://arxiv.org/abs/2506.07969.
  26. [27]Maximilian Herde, Bogdan Raonic, Tobias Rohner, Roger Käppeli, Roberto Molinaro, Emmanuel de Bezenac, and Siddhartha Mishra. Poseidon: Efficient foundation models for PDEs. In The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024. URL https://openreview.net/forum?id=JC1VKK3UXk.arxiv.org/abs/2405.19101
  27. [28]Masanobu Horie and NAOTO MITSUMSE. Physics-embedded neural networks: Graph neural PDE solvers with mixed boundary conditions. In NeurIPS 2022 AI for Science: Progress and Promises, 2022. URL https://openreview.net/forum?id=gyIRLZPGiG.arxiv.org/abs/2205.11912
  28. [29]Peiyan Hu, Haodong Feng, Hongyuan Liu, Tongtong Yan, Wenhao Deng, Tianrun Gao, Rong Zheng, Haoren Zheng, Chenglei Yu, Chuanrui Wang, Kaiwen Li, Zhi-Ming Ma, Dezhi Zhou, Xingcai Lu, Dixia Fan, and Tailin Wu. Readbench: A benchmark for complex physical systems with real-world data. In The Fourteenth International Conference on Learning Representations, 2026. URL https://openreview.net/forum?id=y3oHMcoItR. Oral Presentation.arxiv.org/abs/2601.01829
  29. [30]Qian Huang, Jian Vora, Percy Liang, and Jure Leskovec. Mlagentbench: Evaluating language agents on machine learning experimentation. In International Conference on Machine Learning, 2023. URL https://api.semanticscholar.org/CorpusID:263671541.arxiv.org/abs/2310.03302
  30. [31]Carlos E Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik R Narasimhan. SWE-bench: Can language models resolve real-world github issues? In The Twelfth International Conference on Learning Representations, 2024. URL https://openreview.net/forum?id=VTF8yNQM66.arxiv.org/abs/2310.06770
  31. [32]Felix Koehler, Simon Niedermayr, Rüdiger Westermann, and Nils Thuerey. Apebench: A benchmark for autoregressive neural emulators of pdes. Advances in Neural Information Processing Systems, 37:120252–120310, 2024.DOI
  32. [33]Nikola Kovachki, Zongyi Li, Burigede Liu, Kamyar Azizadehnesheli, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Neural operator: Learning maps between function spaces with applications to pdes. JMLR, 24(1), 2023.
  33. [34]Ruochen Li, Teerth Patel, Qingyun Wang, Qingyun Wang, and Xinya Du. Mlr-copilot: Autonomous machine learning research based on large language models and agents. ArXiv, abs/2408.14033, 2024a. URL https://api.semanticscholar.org/CorpusID:271957477.arxiv.org/abs/2408.14033
  34. [35]Ruoyan Li, Zijie Huang, Haixin Wang, Guancheng Wan, Yizhou Sun, and Wei Wang. Self-guided diffusion model for accelerating computational fluid dynamics, 2025a. URL https://arxiv.org/abs/2504.04375.
  35. [36]Ruoyan Li, Guancheng Wan, Zijie Huang, Zixiao Liu, Haixin Wang, Xiao Luo, Wei Wang, and Yizhou Sun. Flow field reconstruction with sensor placement policy learning. In The Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025b. URL https://openreview.net/forum?id=1esFEjMUBS.
  36. [37]Shanda Li, Tanya Marwah, Junhong Shen, Weiwei Sun, Andrej Risteski, Yiming Yang, and Ameet Talwalkar. Codepde: An inference framework for llm-driven pde solver generation. arXiv preprint arXiv:2505.08783, 2025c.arxiv.org/abs/2505.08783
  37. [38]Zhihao Li, Haoze Song, Di Xiao, Zhilu Lai, and Wei Wang. Harnessing scale and physics: A multi-graph neural operator framework for pdes on arbitrary geometries. arXiv preprint arXiv:2411.15178, 2024b.DOI
  38. [39]Zongyi Li, Nikola Kovachki, Kamyar Azizadehnesheli, Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Multipole graph neural operator for parametric partial differential equations, 2020. URL https://arxiv.org/abs/2006.09535.

Paper details

vlnc:W_7w9t4vy4
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents