Pith. sign in

REVIEW 3 major objections 6 minor 85 references

Reinforcement Learning with Verifiable Physics: Post-training LLMs with Continuous Rewards

T0 review · 3 major / 6 minor · reviewed 2026-07-14 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Post-training language models with continuous physics rewards lets one policy write accurate multi-PDE solvers, and a smaller trained model can beat prompting a much larger frontier model.

desk verdict Solid multi-PDE RL post-training paper: continuous physics rewards beat binary validity and can make a small coder model beat frontier prompting on in-distribution solver generation; transfer is real but selective. read the letter →

arxiv 2607.10474 v1 pith:B23UHDYW submitted 2026-07-11 cs.LG cs.AIcs.CE

Pengfei Cai , Utkarsh Utkarsh , Alan Edelman , Christopher Vincent Rackauckas , Rafael Gomez-Bombarelli This is my paper
classification cs.LGcs.AIcs.CE
keywords reinforcementlearningPDEsolverscodegenerationverifiablerewardslargelanguagemodelsscientificcomputingpost-trainingnumericalmethods
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Building a reliable numerical PDE solver still requires expert choices about discretizations, stability limits, and boundary handling—choices that can make two programs that both run differ by orders of magnitude in solution quality. This paper claims that those graded scientific preferences can be written into a language model’s parameters, not only recovered at inference time with prompting and debugging loops. The method, RLVP, first warms up a model on a bank of teacher solvers, then applies reinforcement learning with a hybrid verifier: hard checks require the code to execute and return finite, correctly shaped fields, while continuous rewards score trajectory accuracy against hidden references and residual consistency with the PDE. One policy trained jointly across eight equation families improves pass rates and solution error over base and supervised baselines, shows selective zero-shot gains on held-out PDEs by recombining numerical motifs, and allows a smaller post-trained model to outperform direct prompting of a frontier model on in-distribution solver generation. If the claim holds, scientific code agents can amortize numerical reliability into the model rather than relying only on larger static models or per-problem scaffolding.

What carries the argument

The hybrid RLVP reward R = V · R_traj · R_phys. Validity V is a hard gate (executes, correct shape, finite values). Continuous factors are expected pass probabilities under exponential tolerances on normalized space-time L2 trajectory error and reference-relative residual error. After supervised fine-tuning on a multi-PDE solver bank, GRPO uses group-relative advantages from this reward to shift probability toward programs that are both executable and physically accurate.

What would settle it

Run the same GRPO setup with validity-only reward versus full continuous physics reward under matched seeds and rollouts: if validity-only matches or beats full RLVP on median best-of-8 nRMSE and held-out transfer, continuous physics is not the driver; or re-evaluate RLVP solvers on new grids, parameters, and independent reference solvers outside the paper’s suite—if high training scores collapse while validity stays high, the signal is overfit to the hidden references.

Watch

Extended reading notes

Core claim

The paper establishes that Reinforcement Learning with Verifiable Physics (RLVP)—post-training a single language-model policy with a hybrid binary–continuous verifier that multiplies program validity by function-space trajectory accuracy and PDE-residual consistency—substantially improves multi-PDE solver code generation. Across model scales, RLVP raises pass@1 and pass@8 and drives median best-of-8 nRMSE to about 6×10^{-4} on eight seen families spanning hyperbolic, parabolic, elliptic, and incompressible-flow systems; a smaller RLVP-trained model outperforms prompting a frontier model under the same direct-generation protocol; and the policy shows selective zero-shot transfer to held-out P

Load-bearing premise

The method assumes that continuous scores against hidden numerical reference trajectories and residuals, with fixed exponential temperatures and a fixed success error threshold, measure true physical correctness rather than closeness to particular reference solvers, grids, and operator combinations.

Editorial extensions

If this is right

  • A single post-trained policy can serve multiple PDE families instead of specializing one model per equation.
  • Continuous physical accuracy rewards improve solution quality and cross-PDE transfer beyond binary execution feedback alone.
  • Smaller verifier-trained models can outperform larger static frontier models on in-distribution PDE solver generation under direct generation.
  • Trained policies recombine numerical motifs (stencils, CFL substepping, RK schemes, boundary handling) into solvers for related unseen PDEs.
  • Future solver-writing systems should put more weight on parameter updates from verifiable physics feedback, not only on inference-time prompting and debugging.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same hybrid continuous-verifier pattern may transfer to other scientific code tasks that admit graded residual or conservation-error signals, such as molecular-dynamics integrators or finite-element assembly.
  • If motif recombination is the transfer mechanism, expanding the solver bank with more operator types could widen zero-shot coverage without a proportional increase in training PDE families.
  • Reference-relative rewards may under-reward stylistically different solvers that still satisfy the PDE better than the hidden reference on new instances.
  • Combining RLVP with retrieval, debugging loops, or verifier-guided search—settings the paper leaves for future work—could compound the amortized gains isolated under direct generation.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper introduces RLVP, a GRPO-based post-training method that adapts LLMs to generate numerical PDE solver code using a hybrid verifier: binary execution validity gates multiplied by continuous rewards for function-space trajectory error (R_traj) and reference-relative residual consistency (R_phys). After SFT on a multi-scheme solver bank spanning eight PDE families (hyperbolic, parabolic, elliptic, incompressible flow), a single policy is trained jointly and evaluated by pass@k and best-of-k nRMSE against hidden numerical references. Across 3B/7B/14B Qwen2.5-Coder models, RLVP substantially improves over base and SFT checkpoints on seen PDEs, can outperform static prompting of larger frontier models under the same direct-generation protocol, and yields selective gains on held-out PDEs that recombine trained numerical motifs. An ablation shows that adding R_traj improves accuracy beyond validity-only RL.

Significance. If the results hold under broader scrutiny, this is a clear and timely contribution: it moves scientific code generation from inference-time scaffolding to amortized post-training with graded physical feedback, and it is among the first multi-PDE RLVR pipelines with continuous rather than binary verifiers. Strengths include a carefully documented solver bank with self-convergence and MMS checks, controlled Base/SFT/RLVP comparisons at three scales, frontier baselines under a matched protocol, a validity-only ablation, and an intended open release of the verifier environment and recipes. The practical claim that a smaller post-trained model can beat frontier prompting on in-distribution solver generation is well supported and of real interest to scientific computing and LLM post-training communities.

major comments (3)
  1. §3.5 Eqs. (3)–(5) and Appendix A fix T_traj=0.05, T_phys=2.0, and the success threshold nRMSE≤10^{-2} without sensitivity analysis. These free parameters define both the dense training signal and the pass@k metric. Because the central claim is that continuous physics rewards (not just validity) drive gains, the paper should report at least a limited sweep or robustness check on T_traj (and preferably T_phys and the success threshold) for the 7B model on seen and held-out aggregates; otherwise it is hard to know how much of Table 1 and Figs. 8–9 depends on this particular temperature/threshold choice.
  2. §4.3 and Appendix F present compositionality and zero-shot transfer as recombination of stencils, CFL control, and time-stepping motifs. The quantitative gains on held-out PDEs are real, but the strongest cases (heat ⊂ reaction–diffusion; 2D advection as a 1D lift; ARD as advection+reaction+diffusion) sit inside the trained operator algebra, while KdV, Darcy–reaction, and advection–diffusion remain weak (Table 16). The abstract and contribution list should more tightly match this selectivity. Please either (i) add a controlled motif-ablation or feature-frequency analysis that separates recombination from generic code quality, or (ii) explicitly frame transfer as selective numerical-motif reuse rather than broad physics internalization.
  3. §3.5 Eq. (1) and Eq. (4): L_traj is measured only against the environment’s fixed numerical references, and R_phys is reference-relative. This is a legitimate RLVR design, but it leaves open whether the policy is learning scheme-agnostic physical correctness or matching reference-solver idiosyncrasies (numerical diffusion, projection residuals, CFL patterns). A load-bearing check for the “verifiable physics” claim would be evaluation under a changed reference family, grid family, or residual discretization not used in training—at least for one or two PDEs—or an explicit limitation stating that rewards are reference-anchored and transfer is within a related numerical-methods neighborhood.
minor comments (6)
  1. Figure 2 and Table 1: clarify whether Claude Sonnet 4.5 and other API baselines used identical system/user prompt constraints (NumPy/SciPy only, fixed interface) as the local models; any prompt mismatch would weaken the “smaller model beats frontier” comparison.
  2. §4.2: Incompressible Navier–Stokes is the clear capacity bottleneck (only 14B RLVP succeeds strongly). A short discussion of why projection structure is harder to learn from the current bank would help readers set expectations.
  3. Related Work: CodePDE and SciML-agent systems are cited; a short explicit contrast on compute amortization (post-training once vs. per-query refinement) would sharpen positioning without new experiments.
  4. Appendix C’s exponential-tolerance derivation is useful; cross-reference it more clearly from §3.5 so readers see that R_traj is an expected-pass probability, not an ad-hoc soft score.
  5. Typos/clarity: “DeepSeek V4 Flash” labeling in Table 1 should be double-checked against the actual API model name; ensure consistent PDE naming (e.g., reaction_diffusion1d vs. Reaction Diffusion 1D) between tables and figures.
  6. Figure 11 (solver component frequencies) is informative; reporting sample sizes and a simple statistical comparison (RLVP vs SFT) would make the motif-reuse narrative less anecdotal.

Circularity Check

0 steps flagged · score 1.0 of 10

No load-bearing circular derivation: RLVP optimizes execution-grounded rewards against hidden numerical references and evaluates on held-out cases/PDEs; standard RLVR, not prediction-by-construction.

full rationale

This paper’s central chain is methodological and empirical, not a first-principles derivation that reduces to its inputs. SFT warm-starts on a curated multi-PDE solver bank; GRPO then updates a policy using R(y,c)=V·R_traj·R_phys (Eq. 5), where V is hard executability and R_traj/R_phys are continuous scores from function-space error and reference-relative residual consistency on hidden instances (Eqs. 1–4; §3.5). Evaluation uses fixed test cases and 10 held-out PDE families never used as SFT targets or RL prompts (§4.1, §4.3). Training on a metric and reporting the same metric on disjoint cases is ordinary RLVR, not self-definitional circularity: the policy is not forced to reproduce a fitted scalar or a quantity defined from the evaluation labels. Appendix C’s exponential-tolerance reading of R_traj is a re-interpretation of the soft reward as an expected binary pass probability, not a claim that a prediction equals a fit. Author self-citations appear only in related SciML context (e.g., physics-constrained generative models) and do not underwrite uniqueness or force the RLVP design. The skeptic concern that dense rewards may overfit reference-solver idiosyncrasies or operator-recombination structure is a generalization/correctness risk, not a circular reduction (no Eq. X ≡ Eq. Y by construction). Mild residual risk only: rewards and success thresholds are defined relative to the same class of numerical references used for grading, so reported nRMSE gains partly measure what was optimized—but held-out PDEs and parameter/IC splits keep this from being tautological. Score 1 reflects that mild coupling without any exhibited circular step.

Assumptions & free parameters 5 free parameters · 5 assumptions · 2 invented entities

The central empirical claim rests on standard RL and numerical-analysis background plus several design choices that are free parameters or paper-specific constructions: exponential tolerance temperatures, success threshold, the multiplicative hybrid reward, and the assumption that hidden numerical references define graded physical correctness for training. No new physical entities are postulated; the invented pieces are methodological (RLVP verifier and multi-PDE solver-bank recipe).

free parameters (5)
  • T_traj (trajectory reward temperature)
    Sets the scale of the exponential expected-pass reward R_traj=exp(-nRMSE/T_traj); fixed at 0.05 without reported sensitivity analysis, yet directly shapes ranking of valid solvers.
  • T_phys (residual reward temperature)
    Temperature for residual-consistency factor; fixed at 2.0 and multiplies into the complete reward, affecting how much residual mismatch attenuates trajectory reward.
  • success_nrmse_threshold
    Binary success for pass@k uses nRMSE≤0.01 for all PDEs; this threshold converts continuous error into the headline reliability metrics.
  • KL coefficient β and GRPO/LoRA optimization hyperparameters
    β=0.001, actor LR 2e-6, LoRA rank 32, 8 rollouts, 50 epochs, etc., are hand-chosen training knobs that the reported gains depend on.
  • Prompt-form sampling weights (0.5/0.35/0.15)
    Mixture of generic, parameter-conditioned, and IC-conditioned prompts is a design choice that shapes the learned policy distribution.
assumptions (5)
  • domain assumption Group-relative policy optimization with token-mean advantages and KL anchor to an SFT reference is a valid post-training method for code-generating policies.
    Invoked throughout §3.6; taken from DeepSeekMath/GRPO literature without re-derivation.
  • domain assumption Normalized L2 trajectory error against a hidden numerical reference is a sufficient proxy for physical solver quality across the studied PDE families.
    Defines R_traj via Eq. (1) and §3.5; standard in PDEBench-style evaluation but still an assumption about what 'correctness' means.
  • ad hoc to paper An exponential distribution over acceptance tolerances yields an appropriate dense RLVR-style pass probability for continuous diagnostics.
    Appendix C motivates this as maximum-entropy on R≥0; other tolerance laws are left to future work, so the reward shape is a modeling choice.
  • domain assumption Teacher solvers in the curated bank are numerically adequate warm-start targets (self-convergence/MMS/cross-checks suffice).
    Appendix B.3 verification protocol underpins the SFT prior; exact closed-form checks are unavailable for many nonlinear cases.
  • standard math Standard finite-difference/volume/spectral numerical analysis results used in the solver bank are reliable background.
    Citations to LeVeque, Toro, Chorin, Trefethen, etc., supply the teacher schemes.
invented entities (2)
  • RLVP hybrid verifier (V · R_traj · R_phys)
    purpose: Gate invalid programs and rank valid PDE solver code by continuous function-space accuracy and residual consistency during RL post-training.
    Core methodological construct of the paper; not a physical entity, but a new reward object the claims depend on.
  • Multi-PDE solver-bank + joint post-training recipe across eight families
    purpose: Provide a shared executable prior and a single policy trained over hyperbolic, parabolic, elliptic, and incompressible-flow systems.
    Dataset/system contribution enabling the multi-PDE claims; evidence is internal experimental, not external prior validation of the bank as a community standard.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reinforcement Learning with Verifiable Physics: Post-training LLMs with Continuous Rewards." pith.science (2026). https://pith.science/paper/B23UHDYW

@misc{pith2026260710474,
  author       = {Pith},
  title        = {Pith review of: Reinforcement Learning with Verifiable Physics: Post-training LLMs with Continuous Rewards},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B23UHDYW}},
  note         = {Machine review of arXiv:2607.10474}
}
read the original abstract

Partial differential equations (PDEs) are foundational to modeling in science and engineering, but constructing reliable numerical solvers remains labor-intensive, demanding expert knowledge of discretization schemes, stability conditions, and boundary treatments. Recent work has begun to frame PDE solving as a code-generation task for large language models (LLMs), yet existing approaches operate primarily at inference time: relying on prompting, debugging, self-refinement, and test-time scaling rather than adapting the model itself. In parallel, reinforcement learning with verifiable rewards has emerged as a post-training paradigm for code and math reasoning, but its verifiers are typically binary: a compiler runs, or a test passes. Such signals discard the graded structure of scientific correctness, where two solvers may both execute and yet differ in solution accuracy by orders of magnitude. In this work, we introduce RLVP: Reinforcement Learning with Verifiable Physics, an RL post-training framework for multi-PDE solver code generation. RLVP addresses this verifiability gap with a hybrid verifier: hard program-validity checks ensure executability, while continuous physics rewards score function-space accuracy and PDE-residual consistency. A single policy is post-trained across diverse PDE families spanning hyperbolic, parabolic, elliptic, and incompressible-flow systems. RLVP improves over both pre-trained and supervised-only baselines on PDE benchmarks, and shows zero-shot improvement transfer to held-out PDEs. We show that a smaller LLM post-trained with RLVP can outperform prompting a frontier model on in-distribution PDE solver generation. The trained policy shows evidence of compositionality in numerical motifs: it recombines stencils, time-stepping schemes, and boundary-handling primitives learned from the PDEs used in training into generated solvers for unseen PDE problems.

Figures

Figures reproduced from arXiv: 2607.10474 by the authors.

Figure 1
Figure 1. Overview of RLVP for multi-PDE solver code generation. We warm-start the policy with supervised fine-tuning (SFT) on a multi-PDE solver bank across diverse PDE problems. During RL, sampled solver programs are executed by the RLVP verifier and scored by a reward combining validity, physical accuracy against hidden numerical references, and physics residual consistency. Group-relative advantages update the policy with… view at source ↗
Figure 2
Figure 2. Progress of evaluation metrics during post-training. We compare Base (Qwen2.5-Coder￾Instruct), SFT, and RLVP post-trained checkpoints on the eight seen PDE tasks. Dashed line: Claude Sonnet 4.5 baseline evaluated with the same k = 8 sampling protocol. Evaluation problems and metrics. We evaluate all models on fixed PDE prompts and case sets. The seen-PDE evaluation uses the same eight PDE families as training, with … view at source ↗
Figure 7
Figure 7. [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figures from the paper (30 more)
Figure 3
Figure 3. Figure 3: Comparison across local and frontier models on seen PDEs. Base, SFT, RLVP, and frontier LLMs are compared by valid execution rate, pass@1, pass@8, and median best-of-8 nRMSE. With RLVP, parameter updates from executable physical feedback can make a smaller policy more …
Figure 4
Figure 4. Figure 4: Evaluation on 10 held-out PDE tasks. We compare Base, SFT, and RLVP checkpoints on unseen PDE prompts not used during SFT or RL training [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Compositional transfer to unseen advection–reaction–diffusion (ARD) (RLVP 7B model). Left: reference and best-of-8 generated solutions for the held-out ARD problem with β = 1, ν = 0.02, and ρ = 2. Right: numerical scheme evidences from generated PDE solver snippets: a …
Figure 6
Figure 6. Figure 6: Self-convergence of the solver bank across the 8 PDE families. Each panel shows the [PITH_FULL_IMAGE:figures/full_fig_p023_6.png]
Figure 7
Figure 7. Figure 7: Training reward over RL steps. RLVP training reward increases smoothly across 3B, 7B, and 14B Qwen2.5-Coder-Instruct models. The curves show that verifier feedback provides a stable optimization signal after the SFT warm start. 25 [PITH_FULL_IMAGE:figures/full_fig_p02…
Figure 8
Figure 8. Figure 8: Ablation of physical accuracy reward vs. validity-only reward for 8 seen PDE tasks. We compare validity-only RL, R = V , against validity plus physical accuracy, R = V Rtraj, for the 7B model. Here, Rtraj is computed from function-space nRMSE against hidden references.…
Figure 9
Figure 9. Figure 9: Ablation of physical accuracy reward vs. validity-only reward for 10 unseen PDE tasks. On the 10 held-out PDE prompts, adding Rtraj to validity improves valid execution rate (0.45 vs. 0.29), pass@1 (0.10 vs. 0.03), pass@8 (0.33 vs. 0.21), and median best-of-8 nRMSE (1.…
Figure 10
Figure 10. Figure 10: Progress of pass@8 evaluation metrics during post-training. Same setup as [PITH_FULL_IMAGE:figures/full_fig_p026_10.png]
Figure 11
Figure 11. Figure 11: Solver component frequencies across model checkpoints. We parse generated solver code for numerical-method features. For solver generation on held-out PDE problems, RLVP post￾trained checkpoints tend to more frequently use periodic boundary logic, np.roll-based stenci…
Figure 12
Figure 12. Figure 12: Dimension lift transfer: 1D advection to 2D advection. The matched seen problem gives the 1D advection solver generated by the RLVP post-trained model. The held-out RLVP solver lifts the transport structure to two dimensions using x/y derivative stencils, directional …
Figure 13
Figure 13. Figure 13: Higher-order synthesis: Cahn–Hilliard. The held-out Cahn–Hilliard PDE prompt requires a fourth-order structure. The solver generated by the RLVP post-trained model constructs the chemical potential µ = −ϵ 2uxx + u 3 − u and applies a second Laplacian, reaching median …
Figure 14
Figure 14. Figure 14: Diffusion-subset transfer: reaction–diffusion to heat. Heat is the diffusion-only subset of the seen reaction–diffusion family during SFT and RL. The solver generated by RLVP post-trained model drops the reaction source while retaining a local periodic Laplacian and e…
Figure 15
Figure 15. Figure 15: Generated solver from the RLVP post-trained model for 1D viscous Burgers. Best selected solver from the 7B RLVP checkpoint on the seen viscous Burgers task. 30 [PITH_FULL_IMAGE:figures/full_fig_p030_15.png]
Figure 16
Figure 16. Figure 16: Generated solver from the RLVP post-trained model for 1D advection. Best selected solver from the 7B RLVP checkpoint on the seen 1D advection task. 31 [PITH_FULL_IMAGE:figures/full_fig_p031_16.png]
Figure 17
Figure 17. Figure 17: Generated solver from the RLVP post-trained model for 1D reaction–diffusion. Best selected solver from the 7B RLVP checkpoint on the seen reaction–diffusion task. 32 [PITH_FULL_IMAGE:figures/full_fig_p032_17.png]
Figure 18
Figure 18. Figure 18: Generated solver from the RLVP post-trained model for 1D diffusion–sorption. Best selected solver from the 7B RLVP checkpoint on the seen diffusion–sorption task. 33 [PITH_FULL_IMAGE:figures/full_fig_p033_18.png]
Figure 19
Figure 19. Figure 19: Generated solver from the RLVP post-trained model for 2D reaction–diffusion. Best selected solver from the 7B RLVP checkpoint on the seen 2D reaction–diffusion task. 34 [PITH_FULL_IMAGE:figures/full_fig_p034_19.png]
Figure 20
Figure 20. Figure 20: Generated solver from the RLVP post-trained model for 2D Darcy flow. Best selected solver from the 7B RLVP checkpoint on the seen Darcy flow task. 35 [PITH_FULL_IMAGE:figures/full_fig_p035_20.png]
Figure 21
Figure 21. Figure 21: Generated solver from the RLVP post-trained model for 2D shallow water equations. Best selected solver from the 7B RLVP checkpoint on the seen shallow water task. 36 [PITH_FULL_IMAGE:figures/full_fig_p036_21.png]
Figure 22
Figure 22. Figure 22: Generated solver from the RLVP post-trained model for 2D incompressible Navier– Stokes. Best selected solver from the 7B RLVP checkpoint on the seen incompressible Navier–Stokes task. 37 [PITH_FULL_IMAGE:figures/full_fig_p037_22.png]
Figure 23
Figure 23. Figure 23: Burgers 1D solutions example for the steeper-multimode initial condition with [PITH_FULL_IMAGE:figures/full_fig_p038_23.png]
Figure 24
Figure 24. Figure 24: Diffusion–sorption 1D solutions example for the uniform initial condition with [PITH_FULL_IMAGE:figures/full_fig_p038_24.png]
Figure 25
Figure 25. Figure 25: Reaction–diffusion 1D solutions example for the front-like initial condition with [PITH_FULL_IMAGE:figures/full_fig_p038_25.png]
Figure 26
Figure 26. Figure 26: Darcy 2D solutions example for the channel diffusion field with [PITH_FULL_IMAGE:figures/full_fig_p038_26.png]
Figure 27
Figure 27. Figure 27: Advection–reaction–diffusion 1D solutions (unseen PDE) example for the front-like initial [PITH_FULL_IMAGE:figures/full_fig_p038_27.png]
Figure 28
Figure 28. Figure 28: Heat 1D solutions (unseen PDE) example for the single-sine initial condition with [PITH_FULL_IMAGE:figures/full_fig_p039_28.png]
Figure 29
Figure 29. Figure 29: Allen–Cahn 2D solutions (unseen PDE) example for the mixed-waves initial condition [PITH_FULL_IMAGE:figures/full_fig_p040_29.png]
Figure 30
Figure 30. Figure 30: Example generic prompt used in the 1D linear advection task. [PITH_FULL_IMAGE:figures/full_fig_p041_30.png]
Figure 31
Figure 31. Figure 31: Example parameter-conditioned prompt used in the 1D reaction–diffusion task. [PITH_FULL_IMAGE:figures/full_fig_p042_31.png]
Figure 32
Figure 32. Figure 32: Example parameter and initial condition conditioned prompt used in the 2D Darcy flow [PITH_FULL_IMAGE:figures/full_fig_p042_32.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

85 extracted references · 2 canonical work pages

  1. [1]

    American Mathematical Society, 2 edition, 2010

    Lawrence C Evans.Partial Differential Equations. American Mathematical Society, 2 edition, 2010

    2010

  2. [2]

    Cambridge university press, 2002

    Randall J LeVeque.Finite volume methods for hyperbolic problems, volume 31. Cambridge university press, 2002

    2002

  3. [3]

    Springer, 1994

    Alfio Quarteroni and Alberto Valli.Numerical approximation of partial differential equations. Springer, 1994

    1994

  4. [4]

    SIAM, 2000

    Lloyd N Trefethen.Spectral methods in MATLAB. SIAM, 2000

    2000

  5. [5]

    SIAM, 1998

    Uri M Ascher and Linda R Petzold.Computer methods for ordinary differential equations and differential-algebraic equations. SIAM, 1998

    1998

  6. [6]

    Evaluating large language models trained on code.arXiv preprint arXiv:2107.03374, 2021

    Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde De Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, et al. Evaluating large language models trained on code.arXiv preprint arXiv:2107.03374, 2021

    arXiv 2021

  7. [7]

    Measuring coding challenge competence with apps.arXiv preprint arXiv:2105.09938, 2021

    Dan Hendrycks, Steven Basart, Saurav Kadavath, Mantas Mazeika, Akul Arora, Ethan Guo, Collin Burns, Samir Puranik, Horace He, Dawn Song, et al. Measuring coding challenge competence with apps.arXiv preprint arXiv:2105.09938, 2021

    arXiv 2021

  8. [8]

    Program synthesis with large language models.arXiv preprint arXiv:2108.07732, 2021

    Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, et al. Program synthesis with large language models.arXiv preprint arXiv:2108.07732, 2021

    arXiv 2021

Show all 85 references
  1. [9]

    Coderl: Mastering code generation through pretrained models and deep reinforcement learning

    Hung Le, Yue Wang, Akhilesh Deepak Gotmare, Silvio Savarese, and Steven Chu Hong Hoi. Coderl: Mastering code generation through pretrained models and deep reinforcement learning. Advances in Neural Information Processing Systems, 35:21314–21328, 2022

    2022

  2. [10]

    Narasimhan, and Yuan Cao

    Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik R. Narasimhan, and Yuan Cao. ReAct: Synergizing reasoning and acting in language models. InThe Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5,

    2023

  3. [11]

    URLhttps://openreview.net/forum?id=WE_vluYUL-X

    OpenReview.net, 2023. URLhttps://openreview.net/forum?id=WE_vluYUL-X

    2023

  4. [12]

    Rlef: Grounding code llms in execution feedback with reinforcement learning.arXiv preprint arXiv:2410.02089, 2024

    Jonas Gehring, Kunhao Zheng, Jade Copet, Vegard Mella, Quentin Carbonneaux, Taco Cohen, and Gabriel Synnaeve. Rlef: Grounding code llms in execution feedback with reinforcement learning.arXiv preprint arXiv:2410.02089, 2024

    2024 arXiv

  5. [13]

    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. Trans. Mach. Learn. Res., 2026. URLhttps://openreview.net/forum?id=eG3Qy5Oux6

    2026

  6. [14]

    Sciml agents: Write the solver, not the solution

    Saarth Gaonkar, Xiang Zheng, Haocheng Xi, Rishabh Tiwari, Kurt Keutzer, Dmitriy Morozov, Michael W Mahoney, and Amir Gholami. Sciml agents: Write the solver, not the solution. arXiv preprint arXiv:2509.09936, 2025

    2025 arXiv

  7. [15]

    Autonumerics: An autonomous, pde-agnostic multi-agent pipeline for scientific computing.arXiv preprint arXiv:2602.17607, 2026

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

    2026

  8. [16]

    All-fem: Agentic large language models fine-tuned for finite element methods.Computer Methods in Applied Mechanics and Engineering, 457:118985, 2026

    Rushikesh Deotale, Adithya Srinivasan, Mahmoud Golestanian, Yuan Tian, Tianyi Zhang, Pavlos Vlachos, and Hector Gomez. All-fem: Agentic large language models fine-tuned for finite element methods.Computer Methods in Applied Mechanics and Engineering, 457:118985, 2026

    2026

  9. [17]

    Pde-agent: A toolchain-augmented multi-agent framework for pde solving.arXiv preprint arXiv:2512.16214, 2025

    Jianming Liu, Ren Zhu, Jian Xu, Kun Ding, Xu-Yao Zhang, Gaofeng Meng, and Cheng-Lin Liu. Pde-agent: A toolchain-augmented multi-agent framework for pde solving.arXiv preprint arXiv:2512.16214, 2025

    2025

  10. [18]

    Pinnsagent: Automated pde surrogation with large language models.arXiv preprint arXiv:2501.12053, 2025

    Qingpo Wuwu, Chonghan Gao, Tianyu Chen, Yihang Huang, Yuekai Zhang, Jianing Wang, Jianxin Li, Haoyi Zhou, and Shanghang Zhang. Pinnsagent: Automated pde surrogation with large language models.arXiv preprint arXiv:2501.12053, 2025. 11

    2025 arXiv

  11. [19]

    Scaling llm test-time compute opti- mally can be more effective than scaling model parameters.arXiv preprint arXiv:2408.03314, 2024

    Charlie Snell, Jaehoon Lee, Kelvin Xu, and Aviral Kumar. Scaling llm test-time compute opti- mally can be more effective than scaling model parameters.arXiv preprint arXiv:2408.03314, 2024

    2024 arXiv

  12. [20]

    Inference scaling laws: An empirical analysis of compute-optimal inference for problem-solving with language models.arXiv preprint arXiv:2408.00724, 2024

    Yangzhen Wu, Zhiqing Sun, Shanda Li, Sean Welleck, and Yiming Yang. Inference scaling laws: An empirical analysis of compute-optimal inference for problem-solving with language models.arXiv preprint arXiv:2408.00724, 2024

    2024 arXiv

  13. [21]

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y . K. Li, Y . Wu, and Daya Guo. DeepSeekMath: Pushing the limits of mathematical reasoning in open language models, 2024. URL https://arxiv.org/abs/ 2402.03300

    2024 arXiv

  14. [22]

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Peiyi Wang, Qihao Zhu, Runxin Xu, Ruoyu Zhang, Shirong Ma, Xiao Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, Aixin Liu, Bing Xue, Bingxuan Wang, Bochao Wu, Bei Feng, Chengda ...

    2025 doi

  15. [23]

    Solving math word problems with process- and outcome-based feedback, 2022

    Jonathan Uesato, Nate Kushman, Ramana Kumar, Francis Song, Noah Siegel, Lisa Wang, Antonia Creswell, Geoffrey Irving, and Irina Higgins. Solving math word problems with process- and outcome-based feedback, 2022. URLhttps://arxiv.org/abs/2211.14275

    2022 arXiv

  16. [24]

    Efficient memory management for large language model serving with pagedattention

    Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph Gonzalez, Hao Zhang, and Ion Stoica. Efficient memory management for large language model serving with pagedattention. InProceedings of the 29th symposium on operating systems principles, pa...

    2023

  17. [25]

    Hybridflow: A flexible and efficient rlhf framework

    Guangming Sheng, Chi Zhang, Zilingfeng Ye, Xibin Wu, Wang Zhang, Ru Zhang, Yanghua Peng, Haibin Lin, and Chuan Wu. Hybridflow: A flexible and efficient rlhf framework. In Proceedings of the Twentieth European Conference on Computer Systems, pages 1279–1297, 2025

    2025

  18. [26]

    Huerta, and Hao Peng

    Minyang Tian, Luyu Gao, Shizhuo Dylan Zhang, Xinan Chen, Cunwei Fan, Xuefei Guo, Roland Haas, Pan Ji, Kittithat Krongchon, Yao Li, Shengyan Liu, Di Luo, Yutao Ma, Hao Tong, Kha Trinh, Chenyu Tian, Zihan Wang, Bohao Wu, Yanyu Xiong, Shengzhu Yin, Minhui Zhu, Kilian 12 Lieret, Y...

    2024 doi

  19. [27]

    Foam-agent: Towards automated intelligent cfd workflows.arXiv preprint arXiv:2505.04997, 2025

    Ling Yue, Nithin Somasekharan, Tingwen Zhang, Yadi Cao, Zhangze Chen, Shimin Di, and Shaowu Pan. Foam-agent: Towards automated intelligent cfd workflows.arXiv preprint arXiv:2505.04997, 2025

    2025

  20. [28]

    Openfoamgpt: A retrieval-augmented large language model (llm) agent for openfoam-based computational fluid dynamics.Physics of Fluids, 37(3), 2025

    Sandeep Pandey, Ran Xu, Wenkang Wang, and Xu Chu. Openfoamgpt: A retrieval-augmented large language model (llm) agent for openfoam-based computational fluid dynamics.Physics of Fluids, 37(3), 2025

    2025

  21. [29]

    Metaopenfoam: an llm-based multi-agent framework for cfd.arXiv preprint arXiv:2407.21320, 2024

    Yuxuan Chen, Xu Zhu, Hua Zhou, and Zhuyin Ren. Metaopenfoam: an llm-based multi-agent framework for cfd.arXiv preprint arXiv:2407.21320, 2024

    2024 arXiv

  22. [30]

    Mooseagent: A llm based multi-agent framework for automating moose simulation, 2025

    Tao Zhang, Zhenhai Liu, Yong Xin, and Yongjun Jiao. Mooseagent: A llm based multi-agent framework for automating moose simulation, 2025. URL https://arxiv.org/abs/2504. 08621

    2025

  23. [31]

    Chronollm: customizing language models for physics-based simulation code generation.Multibody System Dynamics, Feb 2026

    Jingquan Wang, Andrew Negrut, Harry Zhang, Khailanii Slaton, Shu Wang, Radu Serban, Jinlong Wu, and Dan Negrut. Chronollm: customizing language models for physics-based simulation code generation.Multibody System Dynamics, Feb 2026. ISSN 1573-272X. doi: 10.1007/s11044-026-10152-x

    2026 doi

  24. [32]

    Lang- pinn: From language to physics-informed neural networks via a multi-agent framework.arXiv preprint arXiv:2510.05158, 2025

    Xin He, Liangliang You, Hongduan Tian, Bo Han, Ivor Tsang, and Yew-Soon Ong. Lang- pinn: From language to physics-informed neural networks via a multi-agent framework.arXiv preprint arXiv:2510.05158, 2025

    2025

  25. [33]

    Feabench: Evaluating language models on multiphysics reasoning ability

    Nayantara Mudur, Hao Cui, Subhashini Venugopalan, Paul Raccuglia, Michael P Brenner, and Peter Norgaard. Feabench: Evaluating language models on multiphysics reasoning ability. arXiv preprint arXiv:2504.06260, 2025

    2025 arXiv

  26. [34]

    Solving physics olympiad via reinforce- ment learning on physics simulators.arXiv preprint arXiv:2604.11805, 2026

    Mihir Prabhudesai, Aryan Satpathy, Yangmin Li, Zheyang Qin, Nikash Bhardwaj, Amir Zadeh, Chuan Li, Katerina Fragkiadaki, and Deepak Pathak. Solving physics olympiad via reinforce- ment learning on physics simulators.arXiv preprint arXiv:2604.11805, 2026

    2026 arXiv

  27. [35]

    Pde- controller: Llms for autoformalization and reasoning of pdes.arXiv preprint arXiv:2502.00963, 2025

    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

    2025 arXiv

  28. [36]

    Agentic scientific simulation: Execution-grounded model construction and reconstruction.arXiv preprint arXiv:2603.00214, 2026

    Knut-Andreas Lie, Olav Møyner, Elling Svee, and Jakob Torben. Agentic scientific simulation: Execution-grounded model construction and reconstruction.arXiv preprint arXiv:2603.00214, 2026

    2026

  29. [37]

    DAPO: An open-source LLM reinforcement learning system at scale

    Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, YuYue, Weinan Dai, Tiantian Fan, Gaohong Liu, Juncai Liu, LingJun Liu, Xin Liu, Haibin Lin, Zhiqi Lin, Bole Ma, Guangming Sheng, Yuxuan Tong, Chi Zhang, Mofan Zhang, Ru Zhang, Wang Zhang, Hang Zhu, Jinhua Zhu, Jiaz...

    2025

  30. [38]

    Group sequence policy optimization, 2025

    Chujie Zheng, Shixuan Liu, Mingze Li, Xiong-Hui Chen, Bowen Yu, Chang Gao, Kai Dang, Yuqiong Liu, Rui Men, An Yang, Jingren Zhou, and Junyang Lin. Group sequence policy optimization, 2025. URLhttps://arxiv.org/abs/2507.18071

    2025 arXiv

  31. [39]

    Stepcoder: improving code generation with reinforcement learning from compiler feedback

    Shihan Dou, Yan Liu, Haoxiang Jia, Enyu Zhou, Limao Xiong, Junjie Shan, Caishuang Huang, Xiao Wang, Xiaoran Fan, Zhiheng Xi, et al. Stepcoder: improving code generation with reinforcement learning from compiler feedback. InProceedings of the 62nd Annual Meeting of the Associat...

    2024

  32. [40]

    Reinforcement learning for machine learning engineering agents, 2025

    Sherry Yang, Joy He-Yueya, and Percy Liang. Reinforcement learning for machine learning engineering agents, 2025. URLhttps://arxiv.org/abs/2509.01684

    2025 arXiv

  33. [41]

    Beyond binary rewards: Training lms to reason about their uncertainty.arXiv preprint arXiv:2507.16806, 2025

    Mehul Damani, Isha Puri, Stewart Slocum, Idan Shenfeld, Leshem Choshen, Yoon Kim, and Jacob Andreas. Beyond binary rewards: Training lms to reason about their uncertainty.arXiv preprint arXiv:2507.16806, 2025

    2025 arXiv

  34. [42]

    Maziar Raissi, Paris Perdikaris, and George E Karniadakis. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations.Journal of Computational physics, 378:686–707, 2019

    2019

  35. [43]

    Fourier neural operator for parametric partial differen- tial equations.arXiv preprint arXiv:2010.08895, 2020

    Zongyi Li, Nikola Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Fourier neural operator for parametric partial differen- tial equations.arXiv preprint arXiv:2010.08895, 2020

    2010 arXiv

  36. [44]

    Learning nonlinear operators via deeponet based on the universal approximation theorem of operators

    Lu Lu, Pengzhan Jin, Guofei Pang, Zhongqiang Zhang, and George Em Karniadakis. Learning nonlinear operators via deeponet based on the universal approximation theorem of operators. Nature machine intelligence, 3(3):218–229, 2021

    2021

  37. [45]

    Towards long rollout of neural operators with local attention and flow matching-inspired correction: An example in frontal polymerization pdes

    Pengfei Cai, Sulin Liu, Qibang Liu, Philippe Geubelle, and Rafael Gomez-Bombarelli. Towards long rollout of neural operators with local attention and flow matching-inspired correction: An example in frontal polymerization pdes. NeurIPS 2024 Workshop on Machine Learning and the...

    2024

  38. [46]

    Pdebench: An extensive benchmark for scientific machine learning.Advances in neural information processing systems, 35:1596–1611, 2022

    Makoto Takamoto, Timothy Praditia, Raphael Leiteritz, Daniel MacKinlay, Francesco Alesiani, Dirk Pflüger, and Mathias Niepert. Pdebench: An extensive benchmark for scientific machine learning.Advances in neural information processing systems, 35:1596–1611, 2022

    2022

  39. [47]

    Diffusionpde: Generative pde-solving under partial observation.Advances in Neural Information Processing Systems, 37: 130291–130323, 2024

    Jiahe Huang, Guandao Yang, Zichen Wang, and Jeong Joon Park. Diffusionpde: Generative pde-solving under partial observation.Advances in Neural Information Processing Systems, 37: 130291–130323, 2024

    2024

  40. [48]

    Kochmann

    Jan-Hendrik Bastek, WaiChing Sun, and Dennis M. Kochmann. Physics-informed diffusion models, 2024. URLhttps://arxiv.org/abs/2403.14404

    2024 arXiv

  41. [49]

    Physics-constrained flow matching: Sampling generative models with hard constraints.arXiv preprint arXiv:2506.04171, 2025

    Utkarsh Utkarsh, Pengfei Cai, Alan Edelman, Rafael Gomez-Bombarelli, and Christopher Vin- cent Rackauckas. Physics-constrained flow matching: Sampling generative models with hard constraints.arXiv preprint arXiv:2506.04171, 2025

    2025

  42. [50]

    End-to-end probabilistic framework for learning with hard constraints.arXiv preprint arXiv:2506.07003, 2025

    Utkarsh Utkarsh, Danielle C Maddix, Ruijun Ma, Michael W Mahoney, and Yuyang Wang. End-to-end probabilistic framework for learning with hard constraints.arXiv preprint arXiv:2506.07003, 2025

    2025

  43. [51]

    Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, and...

    2022 arXiv

  44. [52]

    Qwen2.5-Coder technical report.arXiv preprint arXiv:2409.12186, 2024

    Binyuan Hui, Jian Yang, Zeyu Cui, Jiaxi Yang, Dayiheng Liu, Lei Zhang, Tianyu Liu, Jiajun Zhang, Bowen Yu, Keming Lu, Kai Dang, Yang Fan, Yichang Zhang, An Yang, Rui Men, Fei Huang, Bo Zheng, Yibo Miao, Shanghaoran Quan, Yunlong Feng, Xingzhang Ren, Xu- ancheng Ren, Jingren Zh...

    2024 arXiv

  45. [53]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language models. 2021. URLhttps://arxiv.org/abs/2106.09685

    2021 arXiv

  46. [54]

    A model for fast computer simulation of waves in excitable media.Physica D: Nonlinear Phenomena, 49(1-2):61–70, 1991

    Dwight Barkley. A model for fast computer simulation of waves in excitable media.Physica D: Nonlinear Phenomena, 49(1-2):61–70, 1991. 14

    1991

  47. [55]

    Finite difference method for numerical computation of discontinuous solutions of the equations of fluid dynamics.Matematiˇ ceskij sbornik, 47(3): 271–306, 1959

    Sergei K Godunov and Ihor Bohachevsky. Finite difference method for numerical computation of discontinuous solutions of the equations of fluid dynamics.Matematiˇ ceskij sbornik, 47(3): 271–306, 1959

    1959

  48. [56]

    Towards the ultimate conservative difference scheme

    Bram Van Leer. Towards the ultimate conservative difference scheme. v. a second-order sequel to godunov’s method.Journal of computational Physics, 32(1):101–136, 1979

    1979

  49. [57]

    New high-resolution central schemes for nonlinear conservation laws and convection–diffusion equations.Journal of computational physics, 160 (1):241–282, 2000

    Alexander Kurganov and Eitan Tadmor. New high-resolution central schemes for nonlinear conservation laws and convection–diffusion equations.Journal of computational physics, 160 (1):241–282, 2000

    2000

  50. [58]

    Toro.Riemann Solvers and Numerical Methods for Fluid Dynamics

    Eleuterio F. Toro.Riemann Solvers and Numerical Methods for Fluid Dynamics. Springer Berlin, Heidelberg, 3 edition, 2009. ISBN 978-3-540-25202-3. doi: 10.1007/b79761

    2009 doi

  51. [59]

    Strong stability-preserving high-order time discretization methods.SIAM review, 43(1):89–112, 2001

    Sigal Gottlieb, Chi-Wang Shu, and Eitan Tadmor. Strong stability-preserving high-order time discretization methods.SIAM review, 43(1):89–112, 2001

    2001

  52. [60]

    On the construction and comparison of difference schemes.SIAM journal on numerical analysis, 5(3):506–517, 1968

    Gilbert Strang. On the construction and comparison of difference schemes.SIAM journal on numerical analysis, 5(3):506–517, 1968

    1968

  53. [61]

    Implicit-explicit runge-kutta methods for time-dependent partial differential equations.Applied Numerical Mathematics, 25(2-3): 151–167, 1997

    Uri M Ascher, Steven J Ruuth, and Raymond J Spiteri. Implicit-explicit runge-kutta methods for time-dependent partial differential equations.Applied Numerical Mathematics, 25(2-3): 151–167, 1997

    1997

  54. [62]

    Fourth-order time-stepping for stiff pdes.SIAM Journal on Scientific Computing, 26(4):1214–1233, 2005

    Aly-Khan Kassam and Lloyd N Trefethen. Fourth-order time-stepping for stiff pdes.SIAM Journal on Scientific Computing, 26(4):1214–1233, 2005

    2005

  55. [63]

    The numerical solution of the Navier–Stokes equations for an incom- pressible fluid.Bulletin of the American Mathematical Society, 73(6):928–931, 1967

    Alexandre Joel Chorin. The numerical solution of the Navier–Stokes equations for an incom- pressible fluid.Bulletin of the American Mathematical Society, 73(6):928–931, 1967. doi: 10.1090/S0002-9904-1967-11853-6

    1967 doi

  56. [64]

    Wellesley-Cambridge Press, 1986

    Gilbert Strang.Introduction to Applied Mathematics. Wellesley-Cambridge Press, 1986

    1986

  57. [65]

    Scipy 1.0: fundamental algorithms for scientific computing in python.Nature methods, 17(3):261–272, 2020

    Pauli Virtanen, Ralf Gommers, Travis E Oliphant, Matt Haberland, Tyler Reddy, David Courna- peau, Evgeni Burovski, Pearu Peterson, Warren Weckesser, Jonathan Bright, et al. Scipy 1.0: fundamental algorithms for scientific computing in python.Nature methods, 17(3):261–272, 2020

    2020

  58. [66]

    On the elimination of aliasing in finite-difference schemes by filtering high-wavenumber components.Journal of Atmospheric Sciences, 28(6):1074–1074, 1971

    Steven A Orszag. On the elimination of aliasing in finite-difference schemes by filtering high-wavenumber components.Journal of Atmospheric Sciences, 28(6):1074–1074, 1971

    1971

  59. [67]

    Semi-lagrangian integration schemes for atmospheric models—a review.Monthly weather review, 119(9):2206–2223, 1991

    Andrew Staniforth and Jean Côté. Semi-lagrangian integration schemes for atmospheric models—a review.Monthly weather review, 119(9):2206–2223, 1991

    1991

  60. [68]

    SIAM, 2007

    Randall J LeVeque.Finite difference methods for ordinary and partial differential equations: steady-state and time-dependent problems. SIAM, 2007

    2007

  61. [69]

    The calculation of the interaction of non-stationary shock waves and obstacles.USSR Computational Mathematics and Mathematical Physics, 1(2):304–320, 1962

    Vladimir Vasil’evich Rusanov. The calculation of the interaction of non-stationary shock waves and obstacles.USSR Computational Mathematics and Mathematical Physics, 1(2):304–320, 1962

    1962

  62. [70]

    Systems of conservation laws.Communications on Pure and Applied Mathematics, 13:217–237, 1960

    Peter Lax and Burton Wendroff. Systems of conservation laws.Communications on Pure and Applied Mathematics, 13:217–237, 1960. doi: 10.1002/cpa.3160130205

    1960 doi

  63. [71]

    The effect of viscosity in hypervelocity impact cratering.Journal of spacecraft and rockets, 40(5):757–763, 2003

    Robert W MacCormack. The effect of viscosity in hypervelocity impact cratering.Journal of spacecraft and rockets, 40(5):757–763, 2003

    2003

  64. [72]

    Springer, 2003

    Willem H Hundsdorfer, Jan G Verwer, and WH Hundsdorfer.Numerical solution of time- dependent advection-diffusion-reaction equations, volume 33. Springer, 2003

    2003

  65. [73]

    Finite volume methods.Handbook of numerical analysis, 7:713–1018, 2000

    Robert Eymard, Thierry Gallouët, and Raphaèle Herbin. Finite volume methods.Handbook of numerical analysis, 7:713–1018, 2000

    2000

  66. [74]

    Methods of conjugate gradients for solving linear systems.Journal of research of the National Bureau of Standards, 49(6):409–436, 1952

    Magnus R Hestenes, Eduard Stiefel, et al. Methods of conjugate gradients for solving linear systems.Journal of research of the National Bureau of Standards, 49(6):409–436, 1952. 15

    1952

  67. [75]

    SIAM, 2003

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

    2003

  68. [76]

    Springer, 4 edition, 2020

    Joel H Ferziger, Milovan Peri´c, and Robert L Street.Computational methods for fluid dynamics. Springer, 4 edition, 2020. doi: 10.1007/978-3-319-99693-6

    2020 doi

  69. [77]

    Approximate riemann solvers, parameter vectors, and difference schemes.Journal of computational physics, 43(2):357–372, 1981

    Philip L Roe. Approximate riemann solvers, parameter vectors, and difference schemes.Journal of computational physics, 43(2):357–372, 1981

    1981

  70. [78]

    Information theory and statistical mechanics.Physical review, 106(4):620, 1957

    Edwin T Jaynes. Information theory and statistical mechanics.Physical review, 106(4):620, 1957

    1957

  71. [79]

    Academic Press, 11 edition, 2014

    Sheldon M Ross.Introduction to Probability Models. Academic Press, 11 edition, 2014

    2014

  72. [80]

    Equation of state calculations by fast computing machines.The journal of chemical physics, 21(6):1087–1092, 1953

    Nicholas Metropolis, Arianna W Rosenbluth, Marshall N Rosenbluth, Augusta H Teller, and Edward Teller. Equation of state calculations by fast computing machines.The journal of chemical physics, 21(6):1087–1092, 1953

    1953

  73. [81]

    Ziebart, Andrew Maas, J

    Brian D. Ziebart, Andrew Maas, J. Andrew Bagnell, and Anind K. Dey. Maximum entropy inverse reinforcement learning. InProceedings of the AAAI Conference on Artificial Intelligence, pages 1433–1438, 2008

    2008

  74. [82]

    Soft actor-critic: Off- policy maximum entropy deep reinforcement learning with a stochastic actor

    Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, and Sergey Levine. Soft actor-critic: Off- policy maximum entropy deep reinforcement learning with a stochastic actor. InInternational conference on machine learning, pages 1861–1870. Pmlr, 2018. 16 Table 2: Hyperparameters for the ...

    2018

  75. [83]

    Where applicable, we additionally require that solvers within a scheme family agree on smooth initial data to within their formal order at fixed resolution

    Self-convergence.Standard refinement triples in N at fixed output times, asserting that errors decrease monotonically and that observed orders match the formal order of the scheme. Where applicable, we additionally require that solvers within a scheme family agree on smooth in...

  76. [84]

    This producesabsoluteerror against the exact ue at each grid level rather than a self-consistency ratio, and detects sign errors and stencil bugs that self- convergence cannot

    Method of manufactured solutions (MMS).For each PDE we construct a smooth an- alytic solution ue, derive the corresponding source term se symbolically, and run each solver on the forced PDE. This producesabsoluteerror against the exact ue at each grid level rather than a self-...

  77. [85]

    This provides a check against an independent benchmark, complementing self-convergence and MMS

    Cross-validation against published reference trajectories.For PDEs covered by PDEBench [45], we run the corresponding selected solvers on the published parameter settings and compare normalized RMSE against the released reference trajectories. This provides a check against an ...

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.