SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2105.02468 · NeurIPS · 2021

Relative stability toward diffeomorphisms indicates performance in deep nets

Mario Geiger, Leonardo Petrini, Alessandro Favero, Matthieu Wyart

arXiv · PDF · Open in the Atlas

Code that ran

3 of 3 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; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 2 failures of our instrument; 0 no recorded run), measured 2026-11-22; 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/2105.02468.json · badge: /api/ran/2105.02468.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
pcsl-epfl/diffeomorphism canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
FunctionStatusWhere it lives
deform Ran
ran; our contract draft was wrong, not the code
pcsl-epfl/diffeomorphism/diff.py
code served (permissive licence) · get_code("8fb1344c81538970")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned Tensor, ndim 2, dtype family float, shape (8, 8) 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 math

torch.manual_seed(0)

# Create a small square image (or batch of images)
# The function expects [..., y, x] where y and x are the same (square)
# Let's use a single 8x8 image
n = 8
image = torch.randn(n, n)

# Temperature parameter
T = 0.5

# High frequency cutoff for the displacement field
cut = 3

# Interpolation method (one of 'linear', 'gaussian', 'nearest')
interp = 'linear'

ARGS = {
    'image': image,
    'T': T,
    'cut': cut,
    'interp': interp
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
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 (8, 8) — shape (8, 8) 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
scalar_field Ran
ran, honoured the contract we drafted
pcsl-epfl/diffeomorphism/diff.py
code served (permissive licence) · get_code("6ad25daacebe0d48")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned Tensor, ndim 2, dtype family float, shape (4, 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 math

torch.manual_seed(0)

# The function scalar_field takes:
# - n: size of the output field (nxn)
# - m: number of modes to use
# - device: device to use (default 'cpu')

# Let's use small values for n and m to keep things tiny
n_val = 4
m_val = 3
device_val = 'cpu'

ARGS = {
    'n': n_val,
    'm': m_val,
    'device': device_val
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
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 (4, 4) — shape (4, 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
scalar_field_modes Ran
ran; our contract draft was wrong, not the code
pcsl-epfl/diffeomorphism/diff.py
code served (permissive licence) · get_code("e57e011fe9cb8615")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) 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
import math

torch.manual_seed(0)

# The function takes:
# - n: number of points in linspace (int)
# - m: number of modes (int)
# - dtype: torch dtype
# - device: device string

# Keep things small as requested
n_val = 4  # number of spatial points
m_val = 3  # number of modes

ARGS = {
    'n': n_val,
    'm': m_val,
    'dtype': torch.float64,
    'device': 'cpu'
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
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

Also present in other papers' code

Of 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 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 2 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
scalar_field_modes 2 other papers
2 later
2112.13547 (2021-12)
PRIME: A few primitives can boost robustness to common corruptions
2209.07735 (2022-09)
Enhance the Visual Representation via Discrete Adversarial Training

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

Understanding why deep nets can classify data in large dimensions remains a challenge. It has been proposed that they do so by becoming stable to diffeomorphisms, yet existing empirical measurements support that it is often not the case. We revisit this question by defining a maximum-entropy distribution on diffeomorphisms, that allows to study typical diffeomorphisms of a given norm. We confirm that stability toward diffeomorphisms does not strongly correlate to performance on benchmark data sets of images. By contrast, we find that the stability toward diffeomorphisms relative to that of generic transformations R f correlates remarkably with the test error t . It is of order unity at initialization but decreases by several decades during training for state-of-the-art architectures. For CIFAR10 and 15 known architectures we find t ≈ 0.2 R f , suggesting that obtaining a small R f is important to achieve good performance. We study how R f depends on the size of the training set and compare it to a simple model of invariant learning. Recently, there has been a considerable effort to understand the benefits of learning features for onehidden-layer fully connected nets. Learning features can occur and improve performance when the 35th Conference on Neural Information Processing Systems (NeurIPS 2021).

For agents

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

get_harvested_code_for_paper("2105.02468")
get_code_for_paper("2105.02468")
have("2105.02468")

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

curl https://syntology.ai/api/ran/2105.02468.json

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

[![Syntology run record](https://syntology.ai/api/ran/2105.02468.svg)](https://syntology.ai/paper/2105.02468)

Connect an agent — have() is free.