This paper introduces COUPLED FLOW, a new neural operator learning framework for multi-physics PDE systems that captures the co-evolution of multiple physical processes by modeling their cross-process interactions. Unlike existing neural operator methods that primarily target single-physics settings, COUPLED FLOW employs a novel structure-constrained flow matching algorithm to learn process- specific mean velocity fields conditioned on the full multi-physics state. This enables fast and accurate one-step prediction of coupled PDE systems that would otherwise require expensive numerical simulations. A key challenge in training COUPLED FLOW however lies in sampling structure-constrained conditional flow interpolants that capture cross-process correlations and serve as training examples for flow matching. To address this challenge, we derive a multi-physics mean-flow consistency identity that characterizes cross-process interactions and defines the structure-constrained mean-flow matching loss. To provide informative training samples for flow matching with this loss, we develop a principled method for sampling correlated conditional interpolants based on posterior multi-output Gaus- sian processes, which capture richer cross-process correlations than the commonly used endpoint-conditioned linear interpolants. We also establish a theoretical re- sult showing that the Wasserstein-2 distance between the predicted and oracle PDE solution distributions is upper-bounded in terms of the structure-constrained flow matching loss and the smoothness of the learned flow. Experiments across diverse coupled PDE benchmarks demonstrate that COUPLED FLOW consistently outperforms existing state-of-the-art neural operators and generative PDE solvers.
Figures & tables
Figure 1: Workflow diagram of the proposed coupled-flow framework which consists of three main components: (a) a structure proposer that constructs Gaussian process-based conditional flow interpolants to capture cross-physics dependencies (Section 3.2 ); (b) a structure filter that constrains coupled transport dynamics through the mean-flow consistency identity (Section 3.1 ); and (c) a one-step inference procedure for efficient multi-physics prediction at testing time (Section 3.3 ).
Method
Lotka-V.
Belousov-Z.
Gray-Scott
Multiphase
THM
FNOc
9.55
0.16
0.55
2.91
0.11
CMWNO
6.65
0.14
0.43
1.93
0.09
UFNOc
8.73
0.17
0.36
3.87
0.08
Transolverc
17.67
1.88
0.49
0.88
0.16
DiffusionPDEc
3.10
0.53
0.21
2.45
0.16
MeanFlowc
13.21
0.17
0.29
3.33
0.13
Table 1: Comparison of relative L2 prediction errors (×10−2) across five multi-physics benchmark datasets with 1024 training samples. Bold and blue values indicate the best and second-best performance, respectively. Detailed results with standard deviations are provided in Table 13 .
Method
Lotka-V.
Belousov-Z.
512
1024
512
1024
MeanFlowc
16.52
13.21
0.17
0.17
CoupledFlow -GP
2.16
1.21
0.11
0.06
CoupledFlow -Linear
2.34
1.29
0.15
0.09
Table 2: Comparison of relative L2 errors (×10−2) on Lotka–Volterra and Belousov–Zhabotinsky. Detailed results with standard deviations are provided in Table 13 of Appendix C .
Method
RD-2D
NS-2D
MeanFlowc
9.51
16.13
FNOc
12.34
9.12
UFNOc
12.33
9.47
COMPOL-RNN
8.28
8.25
COMPOL-ATN
8.27
8.21
CMWNO
8.45
8.42
Table 3: Relative L2 prediction error (×10−2) on Reaction-Diffusion-2D (RD-2D) and Navier-Stokes-2D (NS-2D). Detailed results with standard deviations are provided in Appendix C .
Method
Lotka-V.
Belousov-Z.
Gray-Scott
Multiphase
FNOc
0.70
10.15
0.20
2.04
UFNOc
0.65
10.04
0.25
5.27
CMWNOc
0.62
7.11
0.22
2.36
COMPOL-ATNc
0.48
8.74
0.18
2.67
COMPOL-RNNc
0.52
8.52
0.19
1.87
CViT
0.42
3.70
1.13
3.26
Table 4: Comparison of relative L2 prediction errors (×10−2) with approximately matched parameter counts. The parameter budgets are 10M, 9M, 13M, and 10M for Lotka–Volterra, Belousov–Zhabotinsky, Gray–Scott, and Multiphase, respectively. The best results are highlighted in bold.
Appendix figures & tables15 assets
Supplementary material from the paper’s appendix.
Appendix
Model
Grey-Scot
Multiphase
Lotka-V.
Belousov-Z
THM
Transolver
0.42
0.42
0.42
0.42
0.42
FNO
4.47
4.47
0.58
0.58
8.41
UFNO
6.73
6.73
1.31
1.31
10.04
CMWNO
2.0
2.0
1.0
1.5
5.0
COMPOL-RNN
2.42
2.42
2.50
3.07
10.62
COMPOL-ATN
2.40
2.40
1.49
2.10
10.57
Appendix
Table 5: Number of model parameters (M) of baselines across different benchmark datasets.
Model
Grey-Scot
Multiphase
Belousov-Z
Lotka-V.
THM
FNO
0.278
0.308
0.132
0.142
0.271
UFNO
0.836
0.839
0.492
0.497
0.835
COMPOL-RNN
0.910
0.880
0.608
0.502
1.818
COMPOL-ATN
1.226
1.190
0.661
0.468
2.725
CViT-B
0.539
0.214
0.149
0.208
0.210
CMWNO
11.471
9.449
7.912
11.284
19.870
Appendix
Table 6: The wall clock time (ms) of baselines across different benchmark datasets.
Figure 2: The prediction of coupled-flow -GP (Ours) and other baselines vs. Ground Truth (GT) on Lotka–Volterra across different rollout prediction steps.
Hyperparameter
Grey-Scott
Lotka-V.
Multiphase
Belousov-Z
THM
Architecture
Unet
Unet
Unet
Unet
Unet
# nets
2
2
2
3
5
Params/net
∼ 5.1M
∼ 4.5M
∼ 5.1M
∼ 4.5M
∼ 5.1M
Total params
∼ 10.2M
∼ 9M
∼ 10.2M
∼ 13.5M
∼ 25.5M
Batch size
32
32
32
32
32
Learning rate
6×10−4
6×10−4
6×10−4
6×10−4
6×10−4
Appendix
Table 7: Hyperparameter settings across benchmark datasets. The Adam [ Kingma and Ba, 2017 ] optimizer is referred to as Adam with (β1=0.9,β2=0.999) . SE is referred to as the Squared-Exponential kernel.
Method
MPF-ID
MPF-OOD
DiffusionPDE
0.0241
0.0318
CoupledFlow -Linear
0.0090
0.0240
CoupledFlow -GP
0.0061
0.0190
Appendix
Table 8: Comparison of relative L2 errors under in-distribution (ID) and out-of-distribution (OOD) settings on Multiphase Flow (MPF), using approximately matched model capacity (10M parameters).
Figure 3: Empirical evaluation of the generalization bound in Theorem 2 across five coupled multi-physics benchmarks. We compare the estimated prediction error ERR(θ) (LHS) with its theoretical upper bound κτL(θ) (RHS). The empirical results satisfy the inequality across all datasets, with relatively small differences between the two quantities, suggesting that the bound is informative.
System
Method
Overall L2
BC Error
IC Error
MPF
FNO
0.02687
0.04363
0.02766
CViT
0.14251
0.16921
0.11824
Transolver
0.01325
0.02340
0.01618
COMPOL
0.09545
0.12070
0.09705
CoupledFlow
0.00575
0.00741
0.00539
THM
FNO
0.00198
0.00216
0.00223
Appendix
Table 9: Comparison of overall relative L2 prediction error, boundary condition (BC) error, and initial condition (IC) error on Multiphase Flow (MPF) and Thermal-Hydro-Mechanical (THM) benchmarks using 512 training samples. The best results are highlighted in bold.
Method
Lotka-V.
Belousov-Z.
Gray-Scott
Multiphase
THM
CoupledFlow -Linear
0.0234
0.0015
0.0023
0.0090
0.0011
CoupledFlow -GP
0.0216
0.0011
0.0019
0.0061
0.0006
Sparse GP
0.3554
0.0222
0.1162
0.0781
0.1443
Appendix
Table 10: Comparison of relative L2 prediction errors between CoupledFlow and the sparse Gaussian Process (SGP) baseline across five coupled multi-physics benchmarks. The best results are highlighted in bold.
Method
Lotka-V.
Gray-Scott
coupled-flow -Linear
0.0234 ± 1e-4
0.0023 ± 1e-5
coupled-flow -GP ( 0.25ℓ0 )
0.0226 ± 2e-4
0.0022 ± 3e-5
coupled-flow -GP ( 0.5ℓ0 )
0.0218 ± 8e-5
0.0023 ± 2e-5
coupled-flow -GP ( 1.0ℓ0 )
0.0216 ± 2e-5
0.0019 ± 3e-5
coupled-flow -GP ( 2.0ℓ0 )
0.0215 ± 1e-4
0.0021 ± 4e-5
coupled-flow -GP ( 4.0ℓ0 )
0.0217 ± 5e-5
0.0020 ± 5e-5
Appendix
Table 11: Sensitivity of coupled-flow -GP to the squared-exponential kernel length-scale. We report relative L2 errors ( ×10−2 ) using 512 training samples with ℓ0=1 as the practical implementation.
Kernel
Lotka-V.
Gray-Scott
Our-Linear
0.0234 ± 1e-4
0.0023 ± 1e-5
Squared-Exponential
0.0216 ± 2e-5
0.0019 ± 3e-5
Matérn- 3/2
0.0214 ± 3e-5
0.0022 ± 2e-5
Matérn- 5/2
0.0230 ± 1e-4
0.0023 ± 7e-5
Rational Quadratic
0.0223 ± 2e-5
0.0018 ± 1e-4
Appendix
Table 12: Kernel sensitivity of coupled-flow -GP. We report relative L2 errors (mean ± std) using 512 training samples. All GP kernels use the same default length-scale ℓ0=1 .
Method
Train Samples
Lotka-V.
Belousov-Z.
Grey-Scott
Multiphase
THM
FNOc
512
0.1229 ± 4e-5
0.0023 ± 3e-5
0.0088 ± 4e-5
0.0229 ± 2e-4
0.0015 ± 2e-5
1024
0.0955 ± 6e-5
0.0016 ± 3e-5
0.0055 ± 3e-6
0.0291 ± 1e-5
0.0011 ± 1e-6
CMWNO
512
0.0735 ± 3e-5
0.0021 ± 2e-5
0.0083 ± 5e-4
0.0256 ± 2e-5
0.0013 ± 4e-5
1024
0.0665 ± 1e-5
0.0014 ± 6e-5
0.0043 ± 3e-5
0.0193 ± 6e-5
0.0009 ± 5e-5
UFNOc
512
0.1163 ± 1e-4
0.0026 ± 6e-6
0.0055 ± 5e-5
0.0448 ± 1e-5
0.0013 ± 3e-5
1024
0.0873 ± 8e-5
0.0017 ± 2e-5
0.0036 ± 6e-6
0.0387 ± 1e-5
0.0008 ± 4e-5
Appendix
Table 13: Comparison of relative L2 prediction errors (mean ± std) of our method with different neural operators and generative baselines across different coupled physics datasets. Bold and Blue values indicate the best and second-best performance, respectively.
Method
Time-Step
1 τ
2 τ
5 τ
10 τ
MeanFlowc
0.7420 ± 2e-3
1.3310 ± 5e-3
2.3890 ± 1e-2
4.939 ± 3e-2
FNOc
0.1240 ± 3e-4
0.2610 ± 8e-4
0.4000 ± 1e-3
0.791 ± 2e-3
COMPOL-RNN
0.1090 ± 2e-4
0.2260 ± 5e-4
0.3780 ± 1e-3
0.533 ± 3e-3
COMPOL-ATN
0.1180 ± 2e-4
0.2890 ± 7e-4
0.3640 ± 2e-3
0.580 ± 4e-3
CMWNO
0.0735 ± 1e-4
0.4528 ± 8e-4
0.7176 ± 2e-3
1.030 ± 5e-3
Appendix
Table 14: Comparison of relative L2 error rollout prediction (mean ± std) of our method with different baselines on Lotka-V dataset. Bold and Blue values indicate the best and second-best performance, respectively.
Method
Time-Step
1 τ
2 τ
5 τ
10 τ
MeanFlowc
0.0033 ± 1e-4
0.0102 ± 2e-4
0.0262 ± 7e-4
0.0591 ± 1e-3
FNOc
0.0088 ± 2e-4
0.0139 ± 7e-4
0.0184 ± 1e-3
0.0368 ± 3e-3
COMPOL-RNN
0.0066 ± 1e-4
0.0134 ± 4e-4
0.0121 ± 1e-3
0.0287 ± 2e-3
COMPOL-ATN
0.0051 ± 1e-4
0.0133 ± 5e-4
0.0119 ± 1e-3
0.0211 ± 2e-3
CMWNO
0.0083 ± 2e-4
0.0143 ± 6e-4
0.0196 ± 2e-3
0.0226 ± 4e-3
Appendix
Table 15: Comparison of relative L2 error rollout prediction (mean ± std) of our method with different baselines on GrayScott dataset. Bold and Blue values indicate the best and second-best performance, respectively.
Method
Reaction-Diffusion-2D
Navier-Stokes-2D
MeanFlowc
0.0951 ± 4e-4
0.1613 ± 1e-3
FNOc
0.1234 ± 1e-3
0.0912 ± 7e-4
UFNOc
0.1233 ± 7e-4
0.0947 ± 3e-4
COMPOL-RNN
0.0828 ± 3e-4
0.0825 ± 6e-4
COMPOL-ATN
0.0827 ± 4e-4
0.0821 ± 1e-3
CMWNO
0.0845 ± 7e-4
0.0842 ± 2e-4
Appendix
Table 16: Comparison of relative L2 prediction error (mean ± std) on Reaction-Diffusion-2D and Navier-Stokes-2D. Bold values indicate the best performance for each case. Blue values indicate the second-best performance for each case.
Method
Lotka–Volterra
Gray–Scott
ID coeff.
OOD coeff.
ID coeff.
OOD coeff.
FNO c
0.1229 ± 4e-5
0.1815 ± 1e-4
0.0088 ± 4e-5
0.1015 ± 1e-4
CMWNO
0.0735 ± 3e-5
0.1606 ± 2e-4
0.0083 ± 5e-4
0.1139 ± 3e-4
COMPOL-ATN
0.1132 ± 1e-4
0.1523 ± 3e-4
0.0061 ± 3e-5
0.0888 ± 2e-4
COMPOL-RNN
0.1143 ± 1e-4
0.1627 ± 7e-5
0.0066 ± 1e-4
0.1136 ± 1e-4
DiffusionPDE
0.0441 ± 3e-4
0.1098 ± 1e-4
0.0026 ± 1e-4
0.0067 ± 2e-4
Appendix
Table 17: Coefficient-shift generalization on Lotka–Volterra and Gray–Scott. We report relative L2 errors (mean ± std) under in-distribution (ID) and out-of-distribution (OOD) coefficient settings.