SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.25577 · 2026

Geometric Flow Matching for Molecular Conformation Generation via Manifold Decomposition

Yi Zhou, Wenqi Fan, Yunqing Liu

arXiv · PDF · Open in the Atlas

Code that ran

5 of 7 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 0 have no contract verdict; 5 failures of our instrument; 2 no recorded run), measured 2026-11-22; 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/2605.25577.json · badge: /api/ran/2605.25577.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
DeepGraphLearning/ConfGF canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
FunctionStatusWhere it lives
clip_norm Ran
ran; our contract draft was wrong, not the code
DeepGraphLearning/ConfGF/confgf/utils/torch.py
code served (permissive licence) · get_code("9c0d1fe05c2d642d")
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

torch.manual_seed(0)

# Create a small tensor with shape (batch=2, dim=4)
vec = torch.randn(2, 4)

# limit is a scalar threshold for the norm
limit = 1.5

# p is the norm type (default 2 for L2 norm)
p = 2

ARGS = {
    "vec": vec,
    "limit": limit,
    "p": p,
}
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
embed_3D Ran
ran; our contract draft was wrong, not the code
DeepGraphLearning/ConfGF/confgf/utils/distgeom.py
code served (permissive licence) · get_code("7e7f4a410c3bb886")
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 tuple. 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 torch
torch.manual_seed(0)

# Construct valid arguments for embed_3D

# Number of nodes
num_nodes = 4
# Number of edges
num_edges = 5

# d_target: target distances for each edge
d_target = torch.rand(num_edges) * 2.0 + 0.5  # shape: (num_edges,)

# edge_index: graph connectivity, shape (2, num_edges)
edge_index = torch.tensor([
    [0, 1, 2, 3, 0],
    [1, 2, 3, 0, 2]
], dtype=torch.long)

# init_pos: initial 3D positions, shape (num_nodes, 3)
init_pos = torch.randn(num_nodes, 3)

# edge_order: order of edges (optional, but let's provide it)
edge_order = torch.tensor([1, 1, 2, 2, 3], dtype=torch.long)

# alpha: decay factor for edge weights
alpha = 0.5

# mu: noise scale
mu = 0.1

# step_size: learning rate for optimizer
step_size = 8.0

# num_steps: number of optimization steps
num_steps = 10

# verbose: print loss information
verbose = 1

ARGS = {
    'd_target': d_target,
    'edge_index': edge_index,
    'init_pos': init_pos,
    'edge_order': edge_order,
    'alpha': alpha,
    'mu': mu,
    'step_size': step_size,
    'num_steps': num_steps,
    'verbose': verbose
}
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 tuple — a tuple output has no array shape of its own; none is 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
get_d_from_pos Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
DeepGraphLearning/ConfGF/confgf/utils/distgeom.py
code served (permissive licence) · get_code("188d12cd5c123077")
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 1, dtype family float, shape (3,) 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

torch.manual_seed(0)

# pos: tensor of node positions, shape (num_nodes, num_features)
# edge_index: tensor of edge connections, shape (2, num_edges)
# The function accesses pos[edge_index[0]] and pos[edge_index[1]]
# and calls .norm(dim=-1) on the difference

num_nodes = 2
num_features = 4
num_edges = 3

pos = torch.randn(num_nodes, num_features)
edge_index = torch.tensor([
    [0, 1, 0],  # source nodes
    [1, 0, 1]   # target nodes
], dtype=torch.long)

ARGS = {
    "pos": pos,
    "edge_index": edge_index
}
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 1, dtype family float, shape (3,) — shape (3,) 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
repeat_batch Ran
ran; our contract draft was wrong, not the code
DeepGraphLearning/ConfGF/confgf/utils/torch.py
code served (permissive licence) · get_code("f5f36de2ea7941ea")
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 DataBatch. 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 torch
import copy
from torch_geometric.data import Data, Batch

torch.manual_seed(0)

# Create small graph data objects
data_list = []
for i in range(2):
    x = torch.randn(4, 3)  # 4 nodes, 3 features
    edge_index = torch.tensor([[0, 1, 2], [1, 2, 3]], dtype=torch.long)
    data = Data(x=x, edge_index=edge_index)
    data_list.append(data)

# Create a batch from the data list
batch = Batch.from_data_list(data_list)

num_repeat = 3

ARGS = {
    "batch": batch,
    "num_repeat": num_repeat
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torch_geometric — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataBatch — a DataBatch output has no array shape of its own; none is 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
repeat_data Ran
ran; our contract draft was wrong, not the code
DeepGraphLearning/ConfGF/confgf/utils/torch.py
code served (permissive licence) · get_code("c1f07219e3023a8e")
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 DataBatch. 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 torch
import copy
from torch_geometric.data import Data, Batch

torch.manual_seed(0)

# Create a small graph data object
node_features = torch.randn(4, 3)  # 4 nodes, 3 features
edge_index = torch.tensor([[0, 1, 2, 3], [1, 2, 3, 0]], dtype=torch.long)  # edges
edge_attr = torch.randn(4, 2)  # 4 edges, 2 features
y = torch.tensor([0, 1, 0, 1], dtype=torch.long)  # node labels

data = Data(x=node_features, edge_index=edge_index, edge_attr=edge_attr, y=y)

ARGS = {
    "data": data,
    "num_repeat": 3
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torch_geometric — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataBatch — a DataBatch output has no array shape of its own; none is 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
set_conformer_positions Not yet run
unverified (harvested, no recorded run)
DeepGraphLearning/ConfGF/confgf/utils/chem.py
code served (permissive licence) · get_code("5ad06a60ad46fa2e")
How it ran

no invocation recorded.

update_data_rdmol_positions Not yet run
unverified (harvested, no recorded run)
DeepGraphLearning/ConfGF/confgf/utils/chem.py
code served (permissive licence) · get_code("4738ff7d50f1b6c7")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 4 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 0 later) and 3 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
set_conformer_positions 2 other papers
2 dated earlier
2203.02923 (2022-03)
GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation
2410.20255 (2024-10)
Equivariant Blurring Diffusion for Hierarchical Molecular Conformer Generation
update_data_rdmol_positions 2 other papers
2 dated earlier
2203.02923 (2022-03)
GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation
2410.20255 (2024-10)
Equivariant Blurring Diffusion for Hierarchical Molecular Conformer Generation
repeat_batch 1 other paper
1 dated earlier
2203.02923 (2022-03)
GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation
repeat_data 1 other paper
1 dated earlier
2203.02923 (2022-03)
GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation

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 2026-11-22.

Repositories linked to this paper

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

Abstract

The generation of accurate 3D molecular conformations is a pivotal challenge in computational chemistry and drug discovery. Recently, diffusion and flow matching models have achieved remarkable success. However, there is a critical misalignment between their mathematical formulation and the physical reality of molecules. Existing approaches predominantly treat molecules as unstructured point clouds in Cartesian space, overlooking the intrinsic hierarchical mechanics where bond lengths and bond angles are relatively stiff, whereas torsion angles constitute the dominant flexible degrees of freedom. This lack of manifold awareness forces models to relearn fundamental geometric constraints from scratch, often leading to physically implausible intermediate structures. To address this, we propose GO-Flow that aligns generative modeling with molecular geometry via manifold decomposition. Instead of forcing motion through Euclidean space, GO-Flow decomposes the generation process into three physically motivated subspaces: translation space with linear optimal transport, rotation space with geodesic flows on 𝑆𝑂 (3), and conformation space with entropic optimal transport. This decomposition injects geometric inductive biases and makes the generative paths better aligned with molecular degrees of freedom. When combined with equivariant neural architectures, it encourages rotation-consistent generation and improves geometric validity. Extensive experiments on GEOM-Drugs and GEOM-QM9 demonstrate that GO-Flow achieves state-of-theart generation quality. Notably, by learning naturally straighter probability paths on the correct manifolds, our method enables high-fidelity sampling with as few as 50 steps, effectively bridging the gap between structural precision and computational efficiency. • Applied computing → Molecular structural biology; • Computing methodologies → Neural networks.

For agents

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

get_harvested_code_for_paper("2605.25577")
get_code_for_paper("2605.25577")
have("2605.25577")

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

curl https://syntology.ai/api/ran/2605.25577.json

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

[![Syntology run record](https://syntology.ai/api/ran/2605.25577.svg)](https://syntology.ai/paper/2605.25577)

Connect an agent — have() is free.