SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.01661 · NeurIPS · 2021

Grounding Representation Similarity with Statistical Testing

Jacob Steinhardt, Frances Ding, Jean-Stanislas Denain

arXiv · PDF · Open in the Atlas

Code that ran

7 of 15 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, 2 violated it and 2 have no contract verdict; 3 failures of our instrument; 8 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/2108.01661.json · badge: /api/ran/2108.01661.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
js-d/sim_metric canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
moskomule/anatome community (classifier label)
classifier basis not recorded on the link
2 ran · 6 no recorded run
copy not recorded community
not in the paper's declared repositories
2 ran · 1 no recorded run
brando90/ultimate-anatome extension (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
FunctionStatusWhere it lives
cca_decomp Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
js-d/sim_metric/dists/scoring.py
pointer only (licence: NONE) · get_code("e11bf983b15a4422")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, js-d/sim_metric when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

fft_shift Ran
ran on a synthesized input
moskomule/anatome/anatome/utils.py
code served (permissive licence) · get_code("158351adf6e53105")
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 5, dtype family float, shape (2, 3, 4, 5, 2) 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)

# Create a small FFT-shifted tensor
# The function expects an rFFTed tensor of size [Bx]CxHxWx2
# We'll create a tensor with shape [2, 3, 4, 5, 2] (batch=2, channels=3, height=4, width=5, real/imag=2)
input_tensor = torch.randn(2, 3, 4, 5, 2)

# dims can be None or a tuple of dimensions to shift
dims_value = (2, 3)  # shift over height and width dimensions

ARGS = {
    "input": input_tensor,
    "dims": dims_value,
}
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 5, dtype family float, shape (2, 3, 4, 5, 2) — shape (2, 3, 4, 5, 2) 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
ifft_shift Ran
ran on a synthesized input
moskomule/anatome/anatome/utils.py
code served (permissive licence) · get_code("268e67909741a516")
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 5, dtype family float, shape (2, 3, 4, 5, 2) 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 [B, C, H, W, 2] for rFFT output
# Using distinct sizes: 2 batch, 3 channels, 4 height, 5 width, 2 for real/imag
input_tensor = torch.randn(2, 3, 4, 5, 2)

# dims can be None or a tuple of integers specifying which dimensions to shift
dims_value = (2, 3)  # shift over height and width dimensions

ARGS = {
    "input": input_tensor,
    "dims": dims_value,
}
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 5, dtype family float, shape (2, 3, 4, 5, 2) — shape (2, 3, 4, 5, 2) 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
landscape1d Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from brando90/ultimate-anatome
pointer only · get_code("5193c5e04cb818a8")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

landscape2d Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from brando90/ultimate-anatome
pointer only · get_code("bc2b6fecc0fce1d0")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

mean_cca_corr Ran
ran, violated the contract we drafted
js-d/sim_metric/dists/scoring.py
pointer only (licence: NONE) · get_code("46264c0d57365f61")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, js-d/sim_metric when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

mean_sq_cca_corr Ran
ran, violated the contract we drafted
js-d/sim_metric/dists/scoring.py
pointer only (licence: NONE) · get_code("e47af62595e70b38")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, js-d/sim_metric when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

add_fourier_noise Not yet run
unverified (harvested, no recorded run)
brando90/ultimate-anatome/anatome/fourier.py
code served (permissive licence) · get_code("ae32dd46639c53d7")
How it ran

no invocation recorded.

add_fourier_noise Not yet run
unverified (harvested, no recorded run)
moskomule/anatome/anatome/fourier.py
code served (permissive licence) · get_code("a2999dfd1ff300c1")
How it ran

no invocation recorded.

cca Not yet run
unverified (harvested, no recorded run)
moskomule/anatome/anatome/distance.py
code served (permissive licence) · get_code("e0631598d35a837f")
How it ran

no invocation recorded.

cca_by_qr Not yet run
unverified (harvested, no recorded run)
moskomule/anatome/anatome/distance.py
code served (permissive licence) · get_code("40a17642cadaf3cd")
How it ran

no invocation recorded.

cca_by_svd Not yet run
unverified (harvested, no recorded run)
moskomule/anatome/anatome/distance.py
code served (permissive licence) · get_code("55baeeb075958544")
How it ran

no invocation recorded.

fourier_map Not yet run
unverified (harvested, no recorded run)
this paper's copy was not recorded; identical code first harvested from brando90/ultimate-anatome
pointer only · get_code("428ae24350e49c39")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

landscape1d Not yet run
unverified (harvested, no recorded run)
moskomule/anatome/anatome/landscape.py
code served (permissive licence) · get_code("d0cf26be2713f45f")
How it ran

no invocation recorded.

landscape2d Not yet run
unverified (harvested, no recorded run)
moskomule/anatome/anatome/landscape.py
code served (permissive licence) · get_code("3d27cb293ed77803")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 4 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 4 later) and 11 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
add_fourier_noise 1 other paper
1 later
2406.15812 (2024-06)
UNCOVERING NONLINEAR CONNECTIONS IN MULTIMODAL REPRESENTATIONS
fourier_map 1 other paper
1 later
2406.15812 (2024-06)
UNCOVERING NONLINEAR CONNECTIONS IN MULTIMODAL REPRESENTATIONS
landscape1d 1 other paper
1 later
2406.15812 (2024-06)
UNCOVERING NONLINEAR CONNECTIONS IN MULTIMODAL REPRESENTATIONS
landscape2d 1 other paper
1 later
2406.15812 (2024-06)
UNCOVERING NONLINEAR CONNECTIONS IN MULTIMODAL REPRESENTATIONS

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

To understand neural network behavior, recent works quantitatively compare different networks' learned representations using canonical correlation analysis (CCA), centered kernel alignment (CKA), and other dissimilarity measures. Unfortunately, these widely used measures often disagree on fundamental observations, such as whether deep networks differing only in random initialization learn similar representations. These disagreements raise the question: which, if any, of these dissimilarity measures should we believe? We provide a framework to ground this question through a concrete test: measures should have sensitivity to changes that affect functional behavior, and specificity against changes that do not. We quantify this through a variety of functional behaviors including probing accuracy and robustness to distribution shift, and examine changes such as varying random initialization and deleting principal components. We find that current metrics exhibit different weaknesses, note that a classical baseline performs surprisingly well, and highlight settings where all metrics appear to fail, thus providing a challenge set for further improvement.

For agents

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

get_harvested_code_for_paper("2108.01661")
get_code_for_paper("2108.01661")
have("2108.01661")

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

curl https://syntology.ai/api/ran/2108.01661.json

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

[![Syntology run record](https://syntology.ai/api/ran/2108.01661.svg)](https://syntology.ai/paper/2108.01661)

Connect an agent — have() is free.