SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2004.13688 · 2020

Variational Integrator Graph Networks for Learning Energy Conserving Dynamical Systems

arXiv · PDF · Open in the Atlas

Code that ran

9 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 9 have no contract verdict; 0 failures of our instrument; 2 no recorded run), measured 2026-11-21; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2004.13688.json · badge: /api/ran/2004.13688.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
shaandesai1/VIGN community (classifier label)
classifier basis not recorded on the link
9 ran · 2 no recorded run
FunctionStatusWhere it lives
create_loss_ops Ran
ran on a synthesized input
shaandesai1/VIGN/models.py
code served (permissive licence) · get_code("8fee4976013bd36a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned EagerTensor, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Create small tensors for true and predicted values
# Using batch size 2, dimension 4
true_tensor = tf.constant([[1.0, 2.0, 3.0, 4.0], [5.0, 6.0, 7.0, 8.0]], dtype=tf.float32)
predicted_tensor = tf.constant([[1.1, 2.1, 2.9, 4.2], [4.9, 6.1, 7.0, 7.8]], dtype=tf.float32)

ARGS = {
    "true": true_tensor,
    "predicted": predicted_tensor
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
heinon_heiles Ran
ran on a synthesized input
shaandesai1/VIGN/data_builder.py
code served (permissive licence) · get_code("f0bce4213dbb458f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import tempfile
import os

np.random.seed(0)

# Based on the function signature and body analysis of spring_particle and mass_spring,
# heinon_heiles likely follows the same pattern: a chaotic system integrator

ARGS = {
    'name': tempfile.mktemp(suffix='.pkl'),
    'num_trajectories': 2,
    'NUM_PARTS': 2,
    'T_max': 1.0,
    'dt': 0.01,
    'sub_sample_rate': 0.1,
    'noise_std': 0.0,
    'seed': 0
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
autograd, numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rk1ng Ran
ran on a synthesized input
shaandesai1/VIGN/new_data_builder.py
code served (permissive licence) · get_code("6959b12bcd6d4e06")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Define dx_dt_fn: a function that takes x_t and returns dx/dt
# The function calls dx_dt_fn(x_t), so it needs to accept a tensor and return a tensor
def dx_dt_fn(x_t):
    # Simple linear dynamics: dx/dt = -0.5 * x
    return -0.5 * x_t

# Define x_t: initial state tensor
# Shape: (batch_size=2, state_dim=4)
x_t = torch.randn(2, 4)

# Define dt: time step
dt = 0.1

ARGS = {
    "dx_dt_fn": dx_dt_fn,
    "x_t": x_t,
    "dt": dt,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rk2 Ran
ran on a synthesized input
shaandesai1/VIGN/utils.py
code served (permissive licence) · get_code("d81d23ea9d6a74e8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Define dx_dt_fn - a callable that takes (x_t, ks, ms, bs, nodes) and returns a tensor
def dx_dt_fn(x_t, ks, ms, bs, nodes):
    # Simple dynamics function that returns a tensor of same shape as x_t
    return torch.randn_like(x_t) * 0.1

# x_t: current state tensor (batch_size=2, features=4)
x_t = torch.randn(2, 4)

# ks: spring constants - a tensor or list
ks = torch.tensor([1.0, 2.0, 3.0])

# ms: masses - a tensor or list
ms = torch.tensor([0.5, 1.0, 1.5])

# dt: time step (scalar)
dt = 0.01

# bs: damping coefficients - a tensor or list
bs = torch.tensor([0.1, 0.2, 0.3])

# nodes: number of nodes or node information - scalar or tensor
nodes = 3

ARGS = {
    'dx_dt_fn': dx_dt_fn,
    'x_t': x_t,
    'ks': ks,
    'ms': ms,
    'dt': dt,
    'bs': bs,
    'nodes': nodes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rk3 Ran
ran on a synthesized input
shaandesai1/VIGN/utils.py
code served (permissive licence) · get_code("a96403c99a41ede4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple dx_dt_fn that takes x_t, ks, ms, bs, nodes and returns a tensor
def dx_dt_fn(x_t, ks, ms, bs, nodes):
    # Simple dynamics function: returns a tensor of same shape as x_t
    # Just a simple linear transformation for testing
    return 0.1 * x_t

# Create small tensors
x_t = torch.randn(2, 4)  # batch size 2, dimension 4
ks = torch.randn(3, 4)   # different size for ks
ms = torch.randn(5, 4)   # different size for ms
bs = torch.randn(2, 6)   # batch-related, different size
nodes = torch.randn(7, 4)  # different size for nodes

dt = 0.01  # small time step

ARGS = {
    'dx_dt_fn': dx_dt_fn,
    'x_t': x_t,
    'ks': ks,
    'ms': ms,
    'dt': dt,
    'bs': bs,
    'nodes': nodes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rk3ng Ran
ran on a synthesized input
shaandesai1/VIGN/new_data_builder.py
code served (permissive licence) · get_code("03eb87969ef29717")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple differential equation function
# It takes x_t (a tensor) and returns dx/dt (a tensor of same shape)
def dx_dt_fn(x_t):
    # Simple linear dynamics: dx/dt = -0.1 * x
    return -0.1 * x_t

# Create initial state: batch size 2, dimension 4
x_t = torch.randn(2, 4)

# Time step
dt = 0.01

ARGS = {
    'dx_dt_fn': dx_dt_fn,
    'x_t': x_t,
    'dt': dt
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rk4 Ran
ran on a synthesized input
shaandesai1/VIGN/utils.py
code served (permissive licence) · get_code("1235761bb975f44c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Define dx_dt_fn: a function that takes (x_t, ks, ms, bs, nodes) and returns a tensor of same shape as x_t
def dx_dt_fn(x_t, ks, ms, bs, nodes):
    # Simple derivative function: returns a tensor of same shape as x_t
    return torch.ones_like(x_t) * 0.1

# x_t: state tensor at time t, shape (batch_size, state_dim)
x_t = torch.randn(2, 4)

# ks: spring constants, can be a tensor or list
ks = torch.tensor([1.0, 1.5, 2.0, 2.5])

# ms: masses, can be a tensor or list
ms = torch.tensor([0.5, 1.0, 1.5, 2.0])

# dt: time step (scalar)
dt = 0.01

# bs: damping coefficients or similar, can be a tensor or list
bs = torch.tensor([0.1, 0.2, 0.3, 0.4])

# nodes: number of nodes or node information
nodes = 4

ARGS = {
    'dx_dt_fn': dx_dt_fn,
    'x_t': x_t,
    'ks': ks,
    'ms': ms,
    'dt': dt,
    'bs': bs,
    'nodes': nodes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rk4ng Ran
ran on a synthesized input
shaandesai1/VIGN/new_data_builder.py
code served (permissive licence) · get_code("819bfd6aff5f5d21")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# dx_dt_fn is a callable that takes a tensor and returns a tensor of the same shape
# It represents the derivative function in an ODE solver
def dx_dt_fn(x):
    # Simple linear dynamics: dx/dt = -x
    return -0.5 * x

# x_t is the current state tensor (batch_size=2, state_dim=4)
x_t = torch.randn(2, 4)

# dt is the time step (scalar float)
dt = 0.01

ARGS = {
    "dx_dt_fn": dx_dt_fn,
    "x_t": x_t,
    "dt": dt,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
spring_particle Ran
ran on a synthesized input
shaandesai1/VIGN/data_builder.py
code served (permissive licence) · get_code("431b27b0dbfbe71e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import pickle
import tempfile
import os

# Set random seed for reproducibility
np.random.seed(42)

ARGS = {
    'name': tempfile.gettempdir() + '/spring_particle_test',
    'num_trajectories': 2,
    'NUM_PARTS': 3,
    'T_max': 1.0,
    'dt': 0.1,
    'sub_sample_rate': 0.2,
    'noise_std': 0.01,
    'seed': 42
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
choose_integrator Not yet run
unverified (harvested, no recorded run)
shaandesai1/VIGN/models.py
code served (permissive licence) · get_code("9e6957eb8e22442b")
How it ran

no invocation recorded.

get_dataset Not yet run
unverified (harvested, no recorded run)
shaandesai1/VIGN/data_builder.py
code served (permissive licence) · get_code("65214ed444514ea8")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 10 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
rk4 1 other paper
1 later
2107.08024 (2021-07)
Port-Hamiltonian Neural Networks for Learning Explicit Time-Dependent Dynamical Systems

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Recent advances show that neural networks embedded with physics-informed priors significantly outperform vanilla neural networks in learning and predicting the long term dynamics of complex physical systems from noisy data. Despite this success, there has only been a limited study on how to optimally combine physics priors to improve predictive performance. To tackle this problem we unpack and generalize recent innovations into individual inductive bias segments. As such, we are able to systematically investigate all possible combinations of inductive biases of which existing methods are a natural subset. Using this framework we introduce Variational Integrator Graph Networks - a novel method that unifies the strengths of existing approaches by combining an energy constraint, high-order symplectic variational integrators, and graph neural networks. We demonstrate, across an extensive ablation, that the proposed unifying framework outperforms existing methods, for data-efficient learning and in predictive accuracy, across both single and many-body problems studied in recent literature. We empirically show that the improvements arise because high order variational integrators combined with a potential energy constraint induce coupled learning of generalized position and momentum updates which can be formalized via the Partitioned Runge-Kutta method.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2004.13688")
get_code_for_paper("2004.13688")
have("2004.13688")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2004.13688.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2004.13688.svg)](https://syntology.ai/paper/2004.13688)

Connect an agent — have() is free.