SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2210.05972 · NeurIPS · 2022

Unsupervised Learning of Equivariant Structure from Sequences

Kenji Fukumizu, Takeru Miyato, Masanori Koyama

arXiv · PDF · Open in the Atlas

Code that ran

9 of 9 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 6 have no contract verdict; 2 failures of our instrument; 0 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/2210.05972.json · badge: /api/ran/2210.05972.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
takerum/meta_sequential_prediction canonical (classifier label)
classifier basis not recorded on the link
9 ran · 0 no recorded run
FunctionStatusWhere it lives
conv_downsample Ran
ran on a synthesized input
takerum/meta_sequential_prediction/models/resblock.py
code served (permissive licence) · get_code("38200f7b20908852")
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 4, dtype family float, shape (2, 8, 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 torch.nn as nn
from torch.nn import functional as F

torch.manual_seed(0)

# Create a small convolutional layer
conv = nn.Conv2d(in_channels=4, out_channels=8, kernel_size=3, padding=1)

# Create a small input tensor: batch_size=2, channels=4, height=8, width=8
x = torch.randn(2, 4, 8, 8)

ARGS = {
    "x": x,
    "conv": conv
}
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 4, dtype family float, shape (2, 8, 4, 4) — shape (2, 8, 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
get_index Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
takerum/meta_sequential_prediction/datasets/small_norb.py
code served (permissive licence) · get_code("342adeaf465617bd")
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 ndarray, ndim 1, dtype family int, shape (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 numpy as np
from collections import OrderedDict

# Set random seed for reproducibility
np.random.seed(0)

# The function expects factors as an np array with shape [6, batch_size]
# where each factor takes integer values within the range specified by _NUM_VALUES_PER_FACTOR
# Note: The docstring says [6, batch_size] but there are only 5 factors in _FACTORS_IN_ORDER
# Based on the loop logic, it should actually be [5, batch_size]

batch_size = 2

# Create factors array with shape [5, batch_size]
# Each row corresponds to a factor in _FACTORS_IN_ORDER
# category: values in range(5)
# instance: values in range(5)
# lighting: values in range(6)
# elevation: values in range(9)
# azimuth: values in range(18)

factors = np.array([
    np.random.randint(0, 5, batch_size),   # category
    np.random.randint(0, 5, batch_size),   # instance
    np.random.randint(0, 6, batch_size),   # lighting
    np.random.randint(0, 9, batch_size),   # elevation
    np.random.randint(0, 18, batch_size),  # azimuth
], dtype=np.int64)

ARGS = {
    'factors': factors
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (2,) — shape (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
get_index Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
takerum/meta_sequential_prediction/datasets/three_dim_shapes.py
code served (permissive licence) · get_code("a57cd22c24266687")
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 ndarray, ndim 1, dtype family int, shape (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 numpy as np
from collections import OrderedDict

# Set seed for reproducibility
np.random.seed(0)

# Create a valid factors array
# Shape: [6, batch_size] where batch_size=2
# Each row i must have values in range [0, _NUM_VALUES_PER_FACTOR[_FACTORS_IN_ORDER[i]])
_NUM_VALUES_PER_FACTOR = OrderedDict({
    'floor_hue': 10,
    'wall_hue': 10,
    'object_hue': 10,
    'scale': 8,
    'shape': 4,
    'orientation': 15
})

_FACTORS_IN_ORDER = ['floor_hue', 'wall_hue', 'object_hue', 'scale', 'shape', 'orientation']

batch_size = 2
factors = np.zeros((6, batch_size), dtype=np.int32)

# Fill each factor row with valid values
for i, factor_name in enumerate(_FACTORS_IN_ORDER):
    max_val = _NUM_VALUES_PER_FACTOR[factor_name]
    factors[i, :] = np.random.randint(0, max_val, size=batch_size)

ARGS = {
    'factors': factors
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (2,) — shape (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
make_diagonal Ran
ran on a synthesized input
takerum/meta_sequential_prediction/utils/laplacian.py
code served (permissive licence) · get_code("aeb211ca8536db16")
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 3, dtype family float, shape (2, 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

torch.manual_seed(0)

# The function make_diagonal takes vecs as input
# It expects a tensor and performs operations:
# 1. Adds a dimension at the end: vecs[..., None]
# 2. Repeats it along the last dimension
# 3. Calls make_identity_like which expects shape[-2] == shape[-1]
# So vecs should be a tensor where after the operations it becomes square in last 2 dims

# Create a small vector tensor: batch of 2, dimension 4
vecs = torch.randn(2, 4)

ARGS = {
    "vecs": vecs
}
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 3, dtype family float, shape (2, 4, 4) — shape (2, 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
make_identity Ran
ran on a synthesized input
takerum/meta_sequential_prediction/utils/laplacian.py
code served (permissive licence) · get_code("ae6e56635c2b25df")
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 (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 numpy as np

torch.manual_seed(0)

ARGS = {
    "N": None,
    "D": 4,
    "device": torch.device("cpu")
}
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 (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
make_identity_like Ran
ran on a synthesized input
takerum/meta_sequential_prediction/utils/laplacian.py
code served (permissive licence) · get_code("5053b3c05376c313")
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 3, dtype family float, shape (2, 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

torch.manual_seed(0)

# Create a tensor A with shape that has matching last two dimensions
# Using shape (2, 4, 4) - batch of 2, with 4x4 matrices
A = torch.randn(2, 4, 4)

ARGS = {
    "A": A
}
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 3, dtype family float, shape (2, 4, 4) — shape (2, 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
positionalencoding2d Ran
ran, honoured the contract we drafted
takerum/meta_sequential_prediction/utils/emb2d.py
pointer only (licence: MIT); the stored copy get_code serves, from wzlxjtu/PositionalEncoding2D, is pointer only (licence: NONE) · get_code("69664111108e8701")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from wzlxjtu/PositionalEncoding2D, and GitHub detected no licence in that repository 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.

simclr Ran
ran on a synthesized input
takerum/meta_sequential_prediction/utils/clr.py
code served (permissive licence) · get_code("b11686650315f59e")
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 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 torch
import torch.nn as nn
import torch.nn.functional as F

torch.manual_seed(0)

# The function takes a list of tensors (zs) and several hyperparameters
# zs: list of tensors, each with shape (n, d) where n is batch size and d is embedding dim
# temperature: float for scaling logits
# normalize: bool for L2 normalization
# loss_type: str for similarity computation type

# Create small tensors for zs
batch_size = 2
embedding_dim = 4
num_views = 2  # simclr typically has 2 views (augmentations)

z1 = torch.randn(batch_size, embedding_dim)
z2 = torch.randn(batch_size, embedding_dim)

ARGS = {
    'zs': [z1, z2],
    'temperature': 0.07,
    'normalize': True,
    'loss_type': 'cossim'
}
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 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
upsample_conv Ran
ran on a synthesized input
takerum/meta_sequential_prediction/models/resblock.py
code served (permissive licence) · get_code("0da12a1241154b15")
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 4, dtype family float, shape (2, 3, 16, 16) 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)

# Create input tensor: batch_size=2, channels=4, height=8, width=8
x = torch.randn(2, 4, 8, 8)

# Create a simple convolutional layer
# Input channels: 4, Output channels: 3, Kernel size: 3x3
conv = nn.Conv2d(in_channels=4, out_channels=3, kernel_size=3, padding=1)

ARGS = {
    'x': x,
    'conv': conv
}
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 4, dtype family float, shape (2, 3, 16, 16) — shape (2, 3, 16, 16) 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

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 2 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 2 later) and 6 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
positionalencoding2d 9 other papers
4 dated earlier, 5 later
1706.03762 (2017-06)
Attention Is All You Need
2008.09105 (2020-08)
Location-Aware Graph Convolutional Networks for Video Question Answering
2106.12102 (2021-06)
LegoFormer: Transformers for Block-by-Block Multi-view 3D Reconstruction
and 6 more not named here
get_index 1 other paper
1 later
2310.17325 (2023-10)
C-Disentanglement: Discovering Causally-Independent Generative Factors under an Inductive Bias of Confounder

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

In this study, we present meta-sequential prediction (MSP), an unsupervised framework to learn the symmetry from the time sequence of length at least three. Our method leverages the stationary property (e.g. constant velocity, constant acceleration) of the time sequence to learn the underlying equivariant structure of the dataset by simply training the encoder-decoder model to be able to predict the future observations. We will demonstrate that, with our framework, the hidden disentangled structure of the dataset naturally emerges as a by-product by applying simultaneous block-diagonalization to the transition operators in the latent space, the procedure which is commonly used in representation theory to decompose the feature-space based on the type of response to group actions. We will showcase our method from both empirical and theoretical perspectives. Our result suggests that finding a simple structured relation and learning a model with extrapolation capability are two sides of the same coin. The code is available at https://github.com/takerum/meta_sequential_prediction. .

For agents

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

get_harvested_code_for_paper("2210.05972")
get_code_for_paper("2210.05972")
have("2210.05972")

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

curl https://syntology.ai/api/ran/2210.05972.json

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

[![Syntology run record](https://syntology.ai/api/ran/2210.05972.svg)](https://syntology.ai/paper/2210.05972)

Connect an agent — have() is free.