pith. sign in

arxiv: 2606.10752 · v1 · pith:WAYW7L3Mnew · submitted 2026-06-09 · 💻 cs.AI

AutoPDE: Reliable Agentic PDE Solving via Explicitly Represented Solver Strategies

Huanshuo Dong , Keyao Zhang , Hong Wang , Zhezheng Hao , Zhiwei Zhuang , Ziyan Liu , Jiacong Wang , Gengyuan Liu
show 1 more author
Xin Jin
This is my paper

Pith reviewed 2026-06-27 13:22 UTC · model grok-4.3

classification 💻 cs.AI
keywords PDE solvingLLM coding agentsnumerical methodssolver strategyexplicit representationagentic workflowadaptive tuning
0
0 comments X

The pith

AutoPDE keeps an explicit solver strategy object separate from code so that numerical decisions can be built first and revised from failure evidence.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that reliable PDE solvers require not only code but an inspectable set of decisions on discretization, stabilization, and solver settings that match the equation structure. Current LLM coding agents generate code directly, leaving those decisions implicit and routing all failure feedback back into code edits. AutoPDE instead maintains the strategy as a standalone object that is constructed in three stages before any code is written and can be updated when a numerical solve fails. On the PDE Agent Bench this separation raises the success rate from the prior best of 40.3 percent to 54.5 percent.

Core claim

AutoPDE maintains the solver strategy as an independent, inspectable object that is built before any code is written and revised using numerical evidence whenever a solve fails. The object is populated in three stages that draw from a library of reusable PDE-solving skills: PDE analysis identifies the equation type and algebraic structure; numerical method selection commits to a discretization, stabilization, and linear solver; and adaptive tuning runs low-cost pilot solves to set resolution and tolerances. This explicit representation allows the strategy to be checked and corrected before code generation and to receive targeted revisions from numerical feedback.

What carries the argument

The explicitly represented solver strategy object, an independent inspectable structure built before code and revised from numerical evidence across PDE analysis, method selection, and adaptive tuning stages.

If this is right

  • Numerical decisions become inspectable before code is generated.
  • Failure signals can be used to revise the underlying strategy rather than only the implementation.
  • A shared library of reusable PDE-solving skills can be applied across different equations.
  • Low-cost pilot solves can calibrate resolution and tolerances under given accuracy and runtime budgets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same separation of plan from implementation could be applied to other scientific computing tasks where implicit choices in generated code are hard to debug.
  • Explicit strategy objects might support human oversight or transfer of successful numerical plans between similar PDEs.
  • If strategy revision proves reliable, future agents could maintain libraries of proven strategy templates rather than regenerating plans from scratch each time.

Load-bearing premise

Routing numerical failure feedback to revisions of the explicit strategy object rather than directly to code edits will produce measurable gains in solve success rate.

What would settle it

An ablation that disables strategy revision and routes all failure feedback only to code edits, then measures whether the pass rate falls back to the 40.3 percent baseline level.

Figures

Figures reproduced from arXiv: 2606.10752 by Gengyuan Liu, Hong Wang, huanshuo dong, Jiacong Wang, Keyao Zhang, Xin Jin, Zhezheng Hao, Zhiwei Zhuang, Ziyan Liu.

Figure 1
Figure 1. Figure 1: Three ways to build a PDE solver. (a) Traditional numerical practice: a human expert applies numerical-analysis knowledge to derive an explicit solver strategy (discretization, stabilization, solver configuration), implements it, and on failure revises the strategy rather than only the code. (b) LLM code / PDE agent: a general code agent writes solver.py directly from the PDE specification; the solver stra… view at source ↗
Figure 2
Figure 2. Figure 2: From variational formulation to executable FEM code, illustrated on the Helmholtz [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Ground truth at t = T = 1 and three explicit numerical strategies on the Zalesak slotted￾disk rotation problem; the dashed cyan circle marks where the disk should sit. The table reports the mesh N, wall-clock time of the time-stepping loop on a single CPU core, and the relative L 1 error against the analytic reference. A full mesh sweep for both algorithms, plotted as (runtime, error) Pareto curves, is giv… view at source ↗
Figure 4
Figure 4. Figure 4: Architecture of AutoPDE. Before code generation, PDE analysis and numerical method [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Péclet-stratified behaviour of five Claude-backbone methods on the 41 convection–diffusion [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: (Runtime, rel. L 1 ) Pareto curves for the two numerical strategies of Section 4, obtained by sweeping the mesh N as listed above. The WENO5 curve lies strictly below the first-order upwind curve at every runtime budget we tried: WENO5 at N = 48 already reaches rel. L 1 ≈ 0.78 in 1.36 s, while the finest upwind run we measured, N = 384, needs 34.4 s to reach only 0.85 (25× the wall-clock time for a worse s… view at source ↗
read the original abstract

Numerical solvers for partial differential equations (PDEs) are core computational tools in science and engineering. Building reliable PDE solvers requires not only executable code, but a numerical solver strategy, a set of decisions about discretization, stabilization, solver configuration, and resolution control, that matches the PDE structure. Recent LLM-based coding agents have begun to reduce the programming burden by generating and debugging solver implementations. However, they typically move directly from a PDE problem to solver code, leaving the solver strategy implicit in implementation details. Feedback from a failed solve is therefore routed back to code edits rather than to the underlying strategy, so numerical decisions remain hard to check before code is generated and hard to revise using numerical evidence when it fails. To address this limitation, we propose AutoPDE, a code agent that maintains the solver strategy as an explicitly represented object throughout the solving process: an independent, inspectable object that is built before any code is written and can be revised, using numerical evidence, whenever a solve fails. AutoPDE builds and maintains this object in three stages, all drawing from a library of reusable PDE-solving skills: PDE analysis identifies the equation type and algebraic structure; numerical method selection chooses a numerical method that matches the analysis result and commits to a discretization, stabilization, and linear solver accordingly; and adaptive tuning runs low-cost pilot solves to calibrate resolution and tolerances under the prescribed accuracy and runtime budget. We evaluate AutoPDE on the PDE Agent Bench, where experimental results show that AutoPDE achieves a pass rate of $54.5%$, improving over the strongest baseline by $14.2$ percentage points.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The paper introduces AutoPDE, a code agent for solving partial differential equations (PDEs) that maintains an explicitly represented solver strategy object. This object is constructed in three stages—PDE analysis, numerical method selection, and adaptive tuning—using a library of reusable skills, before any code is generated. The strategy can be revised based on numerical evidence from failed solves. The authors evaluate it on the PDE Agent Bench and report a pass rate of 54.5%, which is 14.2 percentage points higher than the strongest baseline.

Significance. If the performance improvement can be attributed to the explicit strategy representation and its revision mechanism, this work would offer a valuable contribution to the development of reliable LLM-based agents for scientific computing tasks. By making solver strategies inspectable and separable from code, it addresses a limitation in existing approaches where feedback is applied directly to implementations. The staged pipeline and skill library provide a structured framework that could be extended to other domains requiring numerical expertise.

major comments (3)
  1. [§4 (Evaluation)] The experimental results report a 54.5% pass rate and 14.2 pp improvement without providing details on the PDE Agent Bench (e.g., number and types of PDEs, difficulty distribution), the implementation of baseline agents, the number of independent runs, or any statistical tests for significance. This omission prevents verification that the gain is not due to confounding factors such as differences in prompting or library access.
  2. [§3 (Method)] The central claim relies on routing numerical failure feedback to revisions of the explicit strategy object rather than direct code edits. However, the description of the three-stage pipeline does not include the specific mechanism, decision rules, or pseudocode for how numerical evidence updates the strategy object. Without this, it is unclear how the revision process operates or is validated.
  3. [§4 (Evaluation)] No ablation study is presented that isolates the contribution of maintaining and revising the explicit strategy object while holding other elements (such as the skill library and pilot solves) constant. The performance delta cannot be attributed to this design choice based on the current evaluation.
minor comments (1)
  1. The abstract contains LaTeX markup for the percentage values that should be rendered consistently in the final manuscript.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major point below and will revise the manuscript to incorporate the requested clarifications and additions.

read point-by-point responses

  1. Referee: [§4 (Evaluation)] The experimental results report a 54.5% pass rate and 14.2 pp improvement without providing details on the PDE Agent Bench (e.g., number and types of PDEs, difficulty distribution), the implementation of baseline agents, the number of independent runs, or any statistical tests for significance. This omission prevents verification that the gain is not due to confounding factors such as differences in prompting or library access.

    Authors: We agree that the current experimental section lacks sufficient detail for reproducibility and to rule out confounds. In the revised manuscript we will expand §4 with: the exact composition of PDE Agent Bench (number of problems, PDE types, and difficulty distribution); full implementation details for each baseline (prompting, library access, and any other configuration); the number of independent runs and observed variance; and statistical significance tests (e.g., McNemar or bootstrap intervals) on the 14.2 pp difference. These additions will appear in the main text or a new appendix. revision: yes

  2. Referee: [§3 (Method)] The central claim relies on routing numerical failure feedback to revisions of the explicit strategy object rather than direct code edits. However, the description of the three-stage pipeline does not include the specific mechanism, decision rules, or pseudocode for how numerical evidence updates the strategy object. Without this, it is unclear how the revision process operates or is validated.

    Authors: We acknowledge that the revision mechanism is described only at a high level. We will add a dedicated subsection to §3 that specifies the decision rules (e.g., how convergence failure or accuracy violation triggers changes to discretization order, stabilization technique, or method selection from the skill library) and includes pseudocode for the update loop. This will make the routing of numerical evidence to the strategy object explicit and verifiable. revision: yes

  3. Referee: [§4 (Evaluation)] No ablation study is presented that isolates the contribution of maintaining and revising the explicit strategy object while holding other elements (such as the skill library and pilot solves) constant. The performance delta cannot be attributed to this design choice based on the current evaluation.

    Authors: We agree that an ablation isolating the explicit strategy object is needed to attribute the observed gain. We will add such an ablation to the revised §4, comparing the full system against a variant that routes feedback directly to code while keeping the skill library and pilot solves fixed. Results and controls will be reported in the main text or appendix. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical architecture with no derivation chain or fitted predictions

full rationale

The paper describes an LLM-based agent architecture for PDE solving that maintains an explicit solver strategy object. No mathematical derivations, equations, parameter fitting, or predictions are present. Claims rest on benchmark pass rates (54.5% vs. baseline) rather than any self-referential reduction. No self-citations are load-bearing for uniqueness or ansatz; the work is self-contained as an engineering contribution. No steps match any enumerated circularity pattern.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no information on free parameters, axioms, or invented entities; the ledger is therefore empty.

pith-pipeline@v0.9.1-grok · 5851 in / 1192 out tokens · 22991 ms · 2026-06-27T13:22:32.394336+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

61 extracted references · 11 canonical work pages · 4 internal anchors

  1. [1]

    2012 , publisher=

    Automated solution of differential equations by the finite element method: The FEniCS book , author=. 2012 , publisher=

    2012

  2. [2]

    Archive of numerical software , volume=

    The FEniCS project version 1.5 , author=. Archive of numerical software , volume=

  3. [3]

    ACM Transactions on Mathematical Software (TOMS) , volume=

    Firedrake: automating the finite element method by composing abstractions , author=. ACM Transactions on Mathematical Software (TOMS) , volume=. 2016 , publisher=

    2016

  4. [4]

    II—a general-purpose object-oriented finite element library , author=

    deal. II—a general-purpose object-oriented finite element library , author=. ACM Transactions on Mathematical Software (TOMS) , volume=. 2007 , publisher=

    2007

  5. [5]

    Computers & Fluids , volume=

    A numerical solution of the Navier-Stokes equations using the finite element technique , author=. Computers & Fluids , volume=. 1973 , publisher=

    1973

  6. [6]

    Calcolo , volume=

    A stable finite element for the Stokes equations , author=. Calcolo , volume=. 1984 , publisher=

    1984

  7. [7]

    Mathematical Aspects of Finite Element Methods: Proceedings of the Conference Held in Rome, December 10--12, 1975 , pages=

    A mixed finite element method for 2-nd order elliptic problems , author=. Mathematical Aspects of Finite Element Methods: Proceedings of the Conference Held in Rome, December 10--12, 1975 , pages=. 2006 , organization=

    1975

  8. [8]

    Computer methods in applied mechanics and engineering , volume=

    Streamline upwind/Petrov-Galerkin formulations for convection dominated flows with particular emphasis on the incompressible Navier-Stokes equations , author=. Computer methods in applied mechanics and engineering , volume=. 1982 , publisher=

    1982

  9. [9]

    The Galerkin/least-squares method for advective-diffusive equations , author=

    A new finite element formulation for computational fluid dynamics: VIII. The Galerkin/least-squares method for advective-diffusive equations , author=. Computer methods in applied mechanics and engineering , volume=. 1989 , publisher=

    1989

  10. [10]

    SIAM Journal on scientific and statistical computing , volume=

    GMRES: A generalized minimal residual algorithm for solving nonsymmetric linear systems , author=. SIAM Journal on scientific and statistical computing , volume=. 1986 , publisher=

    1986

  11. [11]

    2003 , publisher=

    Iterative methods for sparse linear systems , author=. 2003 , publisher=

    2003

  12. [12]

    Multigrid methods , pages=

    Algebraic multigrid , author=. Multigrid methods , pages=. 1987 , publisher=

    1987

  13. [13]

    Acta numerica , volume=

    Numerical solution of saddle point problems , author=. Acta numerica , volume=. 2005 , publisher=

    2005

  14. [14]

    2014 , publisher=

    Finite elements and fast iterative solvers: with applications in incompressible fluid dynamics , author=. 2014 , publisher=

    2014

  15. [15]

    2013 , publisher=

    A posteriori error estimation techniques for finite element methods , author=. 2013 , publisher=

    2013

  16. [16]

    Computer methods in applied mechanics and engineering , volume=

    A posteriori error estimation in finite element analysis , author=. Computer methods in applied mechanics and engineering , volume=. 1997 , publisher=

    1997

  17. [17]

    SIAM review , volume=

    The p and h-p versions of the finite element method, basic principles and properties , author=. SIAM review , volume=. 1994 , publisher=

    1994

  18. [18]

    Advances in Neural Information Processing Systems , volume=

    Swe-agent: Agent-computer interfaces enable automated software engineering , author=. Advances in Neural Information Processing Systems , volume=

  19. [19]

    SWE-bench: Can Language Models Resolve Real-World GitHub Issues?

    Swe-bench: Can language models resolve real-world github issues? , author=. arXiv preprint arXiv:2310.06770 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  20. [21]

    arXiv preprint arXiv:2510.25803 , year=

    Mixture-of-Experts Operator Transformer for Large-Scale PDE Pre-Training , author=. arXiv preprint arXiv:2510.25803 , year=

    work page arXiv

  21. [23]

    Journal of computational physics , volume=

    Fully multidimensional flux-corrected transport algorithms for fluids , author=. Journal of computational physics , volume=. 1979 , publisher=

    1979

  22. [24]

    Journal of computational physics , volume=

    Weighted essentially non-oscillatory schemes , author=. Journal of computational physics , volume=. 1994 , publisher=

    1994

  23. [25]

    Journal of computational physics , volume=

    Efficient implementation of weighted ENO schemes , author=. Journal of computational physics , volume=. 1996 , publisher=

    1996

  24. [26]

    The twelfth international conference on learning representations , year=

    MetaGPT: Meta programming for a multi-agent collaborative framework , author=. The twelfth international conference on learning representations , year=

  25. [27]

    Advances in neural information processing systems , volume=

    Self-refine: Iterative refinement with self-feedback , author=. Advances in neural information processing systems , volume=

  26. [32]

    Journal of Computational physics , volume=

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , author=. Journal of Computational physics , volume=. 2019 , publisher=

    2019

  27. [33]

    Fourier Neural Operator for Parametric Partial Differential Equations

    Fourier neural operator for parametric partial differential equations , author=. arXiv preprint arXiv:2010.08895 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  28. [34]

    Advances in Neural Information Processing Systems , volume=

    Poseidon: Efficient foundation models for pdes , author=. Advances in Neural Information Processing Systems , volume=

  29. [35]

    DPOT: Auto-regressive denoising operator transformer for large-scale pde pre-training.arXiv preprint arXiv:2403.03542, 2024

    Dpot: Auto-regressive denoising operator transformer for large-scale pde pre-training , author=. arXiv preprint arXiv:2403.03542 , year=

    work page arXiv

  30. [36]

    arXiv preprint arXiv:2403.07187 , year=

    Ups: Efficiently building foundation models for pde solving via cross-modal adaptation , author=. arXiv preprint arXiv:2403.07187 , year=

    work page arXiv

  31. [37]

    Huang, Zhen and Yashengjiang, Yushan and Li, Junhui and Dong, Huanshuo and Wei, Yangbo and Hao, Zhezheng and Ma, Jiangtao and Bai, Songlin and Hao, Zhongkai and Yue, Xihang and Si, Guangzong and Jiang, Dongming and Yao, Chao and Hu, Zhanhua and Zhang, Jianqing and Liu, Pengwei and Shen, Yaomin and Ren, Xingyu and Liu, Lei and Xu, Zikang and Li, Han and Ya...

  32. [38]

    A posteriori error estimation in finite element analysis

    Mark Ainsworth and J Tinsley Oden. A posteriori error estimation in finite element analysis. Computer methods in applied mechanics and engineering, 142 0 (1-2): 0 1--88, 1997

    1997

  33. [39]

    The fenics project version 1.5

    Martin Aln s, Jan Blechta, Johan Hake, August Johansson, Benjamin Kehlet, Anders Logg, Chris Richardson, Johannes Ring, Marie E Rognes, and Garth N Wells. The fenics project version 1.5. Archive of numerical software, 3 0 (100), 2015

    2015

  34. [40]

    A stable finite element for the stokes equations

    Douglas N Arnold, Franco Brezzi, and Michel Fortin. A stable finite element for the stokes equations. Calcolo, 21 0 (4): 0 337--344, 1984

    1984

  35. [41]

    The p and h-p versions of the finite element method, basic principles and properties

    Ivo Babu s ka and Manil Suri. The p and h-p versions of the finite element method, basic principles and properties. SIAM review, 36 0 (4): 0 578--632, 1994

    1994

  36. [42]

    Wolfgang Bangerth, Ralf Hartmann, and Guido Kanschat. deal. ii—a general-purpose object-oriented finite element library. ACM Transactions on Mathematical Software (TOMS), 33 0 (4): 0 24--es, 2007

    2007

  37. [43]

    Numerical solution of saddle point problems

    Michele Benzi, Gene H Golub, and J \"o rg Liesen. Numerical solution of saddle point problems. Acta numerica, 14: 0 1--137, 2005

    2005

  38. [44]

    Streamline upwind/petrov-galerkin formulations for convection dominated flows with particular emphasis on the incompressible navier-stokes equations

    Alexander N Brooks and Thomas JR Hughes. Streamline upwind/petrov-galerkin formulations for convection dominated flows with particular emphasis on the incompressible navier-stokes equations. Computer methods in applied mechanics and engineering, 32 0 (1-3): 0 199--259, 1982

    1982

  39. [45]

    Teaching Large Language Models to Self-Debug

    Xinyun Chen, Maxwell Lin, Nathanael Sch \"a rli, and Denny Zhou. Teaching large language models to self-debug. arXiv preprint arXiv:2304.05128, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  40. [46]

    AutoNumerics: An au- tonomous, PDE-agnostic multi-agent pipeline for scien- tific computing

    Jianda Du, Youran Sun, and Haizhao Yang. Autonumerics: An autonomous, pde-agnostic multi-agent pipeline for scientific computing. arXiv preprint arXiv:2602.17607, 2026

    work page arXiv 2026

  41. [47]

    Finite elements and fast iterative solvers: with applications in incompressible fluid dynamics

    Howard C Elman, David J Silvester, and Andrew J Wathen. Finite elements and fast iterative solvers: with applications in incompressible fluid dynamics. Oxford university press, 2014

    2014

  42. [48]

    Pde-sharp: Pde solver hybrids through analysis and refinement passes

    Shaghayegh Fazliani and Madeleine Udell. Pde-sharp: Pde solver hybrids through analysis and refinement passes. arXiv preprint arXiv:2511.00183, 2025

    work page arXiv 2025

  43. [49]

    Metagpt: Meta programming for a multi-agent collaborative framework

    Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, et al. Metagpt: Meta programming for a multi-agent collaborative framework. In The twelfth international conference on learning representations, 2023

    2023

  44. [50]

    PDEAgent-Bench : A multi-metric benchmark for pde-to-solver code generation, 2026

    Zhen Huang, Yushan Yashengjiang, Junhui Li, Huanshuo Dong, Yangbo Wei, Zhezheng Hao, Jiangtao Ma, Songlin Bai, Zhongkai Hao, Xihang Yue, Guangzong Si, Dongming Jiang, Chao Yao, Zhanhua Hu, Jianqing 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 benchmark for...

    2026

  45. [51]

    A new finite element formulation for computational fluid dynamics: Viii

    Thomas JR Hughes, Leopoldo P Franca, and Gregory M Hulbert. A new finite element formulation for computational fluid dynamics: Viii. the galerkin/least-squares method for advective-diffusive equations. Computer methods in applied mechanics and engineering, 73 0 (2): 0 173--189, 1989

    1989

  46. [52]

    Efficient implementation of weighted eno schemes

    Guang-Shan Jiang and Chi-Wang Shu. Efficient implementation of weighted eno schemes. Journal of computational physics, 126 0 (1): 0 202--228, 1996

    1996

  47. [53]

    Codepde: An inference framework for llm-driven pde solver generation

    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, 2025

    work page arXiv 2025

  48. [54]

    Weighted essentially non-oscillatory schemes

    Xu-Dong Liu, Stanley Osher, and Tony Chan. Weighted essentially non-oscillatory schemes. Journal of computational physics, 115 0 (1): 0 200--212, 1994

    1994

  49. [55]

    Automated solution of differential equations by the finite element method: The FEniCS book, volume 84

    Anders Logg, Kent-Andre Mardal, and Garth Wells. Automated solution of differential equations by the finite element method: The FEniCS book, volume 84. Springer Science & Business Media, 2012

    2012

  50. [56]

    Self-refine: Iterative refinement with self-feedback

    Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, et al. Self-refine: Iterative refinement with self-feedback. Advances in neural information processing systems, 36: 0 46534--46594, 2023

    2023

  51. [57]

    Firedrake: automating the finite element method by composing abstractions

    Florian Rathgeber, David A Ham, Lawrence Mitchell, Michael Lange, Fabio Luporini, Andrew TT McRae, Gheorghe-Teodor Bercea, Graham R Markall, and Paul HJ Kelly. Firedrake: automating the finite element method by composing abstractions. ACM Transactions on Mathematical Software (TOMS), 43 0 (3): 0 1--27, 2016

    2016

  52. [58]

    A mixed finite element method for 2-nd order elliptic problems

    Pierre-Arnaud Raviart and Jean-Marie Thomas. A mixed finite element method for 2-nd order elliptic problems. In Mathematical Aspects of Finite Element Methods: Proceedings of the Conference Held in Rome, December 10--12, 1975, pages 292--315. Springer, 2006

    1975

  53. [59]

    Algebraic multigrid

    John W Ruge and Klaus St \"u ben. Algebraic multigrid. In Multigrid methods, pages 73--130. SIAM, 1987

    1987

  54. [60]

    Gmres: A generalized minimal residual algorithm for solving nonsymmetric linear systems

    Youcef Saad and Martin H Schultz. Gmres: A generalized minimal residual algorithm for solving nonsymmetric linear systems. SIAM Journal on scientific and statistical computing, 7 0 (3): 0 856--869, 1986

    1986

  55. [61]

    Iterative methods for sparse linear systems

    Yousef Saad. Iterative methods for sparse linear systems. SIAM, 2003

    2003

  56. [62]

    Pde-controller: Llms for autoformalization and reasoning of pdes

    Mauricio Soroco, Jialin Song, Mengzhou Xia, Kye Emond, Weiran Sun, and Wuyang Chen. Pde-controller: Llms for autoformalization and reasoning of pdes. arXiv preprint arXiv:2502.00963, 2025

    work page arXiv 2025

  57. [63]

    A numerical solution of the navier-stokes equations using the finite element technique

    Cedric Taylor and Paul Hood. A numerical solution of the navier-stokes equations using the finite element technique. Computers & Fluids, 1 0 (1): 0 73--100, 1973

    1973

  58. [64]

    u diger Verf \

    R \"u diger Verf \"u rth. A posteriori error estimation techniques for finite element methods. OUP Oxford, 2013

    2013

  59. [65]

    OpenHands: An Open Platform for AI Software Developers as Generalist Agents

    Xingyao Wang, Boxuan Li, Yufan Song, Frank F Xu, Xiangru Tang, Mingchen Zhuge, Jiayi Pan, Yueqi Song, Bowen Li, Jaskirat Singh, et al. Openhands: An open platform for ai software developers as generalist agents. arXiv preprint arXiv:2407.16741, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  60. [66]

    Swe-agent: Agent-computer interfaces enable automated software engineering

    John Yang, Carlos E Jimenez, Alexander Wettig, Kilian Lieret, Shunyu Yao, Karthik Narasimhan, and Ofir Press. Swe-agent: Agent-computer interfaces enable automated software engineering. Advances in Neural Information Processing Systems, 37: 0 50528--50652, 2024

    2024

  61. [67]

    Fully multidimensional flux-corrected transport algorithms for fluids

    Steven T Zalesak. Fully multidimensional flux-corrected transport algorithms for fluids. Journal of computational physics, 31 0 (3): 0 335--362, 1979

    1979