SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2410.05016 · ICLR · 2025

T-JEPA: Augmentation-Free Self-Supervised Learning for Tabular Data

Hugo Thimonier, José De Melo Costa, Fabrice Popineau, Arpad Rimmel, Bich-Liên Doan

arXiv · PDF · Open in the Atlas

Code that ran

10 of 14 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (4 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 2 violated it and 4 have no contract verdict; 3 failures of our instrument; 4 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/2410.05016.json · badge: /api/ran/2410.05016.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
jose-melo/t-jepa community
Papers with Code archive
10 ran · 4 no recorded run
FunctionStatusWhere it lives
PositionalEncoding Ran
ran on a synthesized input · constructed an object; no result was computed
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("088ac74f7c0459d3")
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 an instance of its own class, PositionalEncoding: the run constructed the object and did not call it, so no result was computed. 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
from torch.nn import Dropout, LayerNorm, Linear, Module, ModuleList, MultiheadAttention

torch.manual_seed(0)

ARGS = {
    'd_model': 4,
    'dropout': 0.1,
    'max_len': 5000
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type PositionalEncoding — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a PositionalEncoding 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
Tokenizer Ran
ran on a synthesized input · constructed an object; no result was computed
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("c7213c185d77231d")
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 an instance of its own class, Tokenizer: the run constructed the object and did not call it, so no result was computed. 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 math
import torch
import torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'd_numerical': 4,
    'categories': [3, 5, 2],
    'd_token': 8,
    'bias': True,
    'n_cls_tokens': 1,
}
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 Tokenizer — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Tokenizer 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
TransformerEncoder Ran
ran on a synthesized input · constructed an object; no result was computed
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("f11d932208f0a4ed")
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 an instance of its own class, TransformerEncoder: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch.nn as nn

torch.manual_seed(0)

# Create a minimal TransformerEncoderLayer
d_model = 4
nhead = 2
dim_feedforward = 8
dropout = 0.1

encoder_layer = nn.TransformerEncoderLayer(
    d_model=d_model,
    nhead=nhead,
    dim_feedforward=dim_feedforward,
    dropout=dropout,
    batch_first=True,
    activation='relu'
)

# Create a minimal LayerNorm for the norm parameter
norm = nn.LayerNorm(d_model)

ARGS = {
    'encoder_layer': encoder_layer,
    'num_layers': 2,
    'norm': norm,
    'enable_nested_tensor': True,
    'mask_check': True
}
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 TransformerEncoder — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a TransformerEncoder 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
TransformerEncoderLayer Ran
ran on a synthesized input · constructed an object; no result was computed
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("9a5b310fb0c45b61")
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 an instance of its own class, TransformerEncoderLayer: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch.nn as nn
from torch.nn import Dropout, LayerNorm, Linear, Module, ModuleList, MultiheadAttention
import torch.nn.functional as F

torch.manual_seed(0)

# Build ARGS dictionary for TransformerEncoderLayer
ARGS = {
    'd_model': 64,
    'nhead': 8,
    'dim_feedforward': 256,
    'dropout': 0.1,
    'activation': 'relu',
    'layer_norm_eps': 1e-5,
    'batch_first': False,
    'norm_first': False,
    'bias': True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type TransformerEncoderLayer — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a TransformerEncoderLayer 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
_detect_is_causal_mask Ran
ran, violated the contract we drafted
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("be75e3fe4546c295")
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 bool, 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
from torch import Tensor

torch.manual_seed(0)

# Test case 1: mask is None, is_causal is None
mask_none = None
is_causal_none = None
size_none = None

# Test case 2: mask is a causal mask
sz = 5
causal_mask = torch.triu(torch.full((sz, sz), float('-inf'), dtype=torch.float32, device=torch.device('cpu')), diagonal=1)

# Test case 3: mask is not a causal mask
non_causal_mask = torch.zeros((5, 5), dtype=torch.float32, device=torch.device('cpu'))

# Test case 4: is_causal is explicitly True
is_causal_true = True

# Test case 5: size parameter specified
size_param = 5

# Create multiple test cases
ARGS = [
    {
        "mask": None,
        "is_causal": None,
        "size": None
    },
    {
        "mask": causal_mask,
        "is_causal": None,
        "size": None
    },
    {
        "mask": non_causal_mask,
        "is_causal": None,
        "size": None
    },
    {
        "mask": causal_mask,
        "is_causal": True,
        "size": None
    },
    {
        "mask": causal_mask,
        "is_causal": None,
        "size": 5
    },
]

# Use the first test case as the primary ARGS
ARGS = {
    "mask": causal_mask,
    "is_causal": None,
    "size": None
}
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 bool, 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
_generate_square_subsequent_mask Ran
ran, violated the contract we drafted
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("65aafd77d580c49d")
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 (5, 5) 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)

ARGS = {
    "sz": 5,
    "device": torch.device("cpu"),
    "dtype": torch.float32,
}
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 (5, 5) — shape (5, 5) 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_activation_fn Ran
ran; our contract draft was wrong, not the code
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("7372b44b96fac487")
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 function. 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 torch.nn.functional as F
from torch import Tensor

torch.manual_seed(0)

ARGS = {
    'activation': 'relu'
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
_get_clones Ran
ran; our contract draft was wrong, not the code
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("8487b07e2fbe3aa2")
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) and returned ModuleList. 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.nn import Dropout, LayerNorm, Linear, Module, ModuleList, MultiheadAttention

torch.manual_seed(0)

# Create a simple module to clone
class SimpleModule(Module):
    def __init__(self):
        super().__init__()
        self.linear = Linear(4, 3)
        self.dropout = Dropout(0.1)
    
    def forward(self, x):
        return self.dropout(self.linear(x))

# Instantiate the module
module = SimpleModule()

# Number of clones to create
N = 3

ARGS = {
    "module": module,
    "N": N
}
Setup block drafted by
a language model, Claude Haiku 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 ModuleList — a ModuleList output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ModuleList output the recorded ndim 1 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
_get_seq_len Ran
ran, honoured the contract we drafted
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("2e73a89e42e7fd36")
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 int, ndim 0, dtype family int, 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
from typing import Optional

torch.manual_seed(0)

# Test case 1: 2D tensor (seq_len, batch)
src_2d = torch.randn(5, 2)
batch_first_1 = False

# Test case 2: 3D tensor with batch_first=True (batch, seq_len, features)
src_3d_batch_first = torch.randn(2, 5, 4)
batch_first_2 = True

# Test case 3: 3D tensor with batch_first=False (seq_len, batch, features)
src_3d_seq_first = torch.randn(5, 2, 4)
batch_first_3 = False

# Create ARGS for the function call
# We'll test with the 3D batch_first case as primary
ARGS = {
    "src": src_3d_batch_first,
    "batch_first": batch_first_2
}
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 int, ndim 0, dtype family int, 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
torch_cast_to_dtype Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("c3dd55e49fd28138")
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)

# Create a small tensor to cast
obj = torch.randn(2, 4)

# Test with 'float32' dtype
dtype_name = 'float32'

ARGS = {
    'obj': obj,
    'dtype_name': dtype_name
}
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
Encoder Not yet run
unverified (harvested, no recorded run)
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("f36b97662fdb54f7")
How it ran

no invocation recorded.

TabularEncoder Not yet run
unverified (harvested, no recorded run)
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("ca39498dce297652")
How it ran

no invocation recorded.

_debug_values Not yet run
unverified (harvested, no recorded run)
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("3ef6c6b262613f76")
How it ran

no invocation recorded.

apply_masks_from_idx Not yet run
unverified (harvested, no recorded run)
jose-melo/t-jepa/src/encoder.py
code served (permissive licence) · get_code("a2d9ad216146c0fd")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 3 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 1 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
_get_clones
brick of this paper's assembly: helper
6 other papers
6 dated earlier · 2 only as an assembly brick
2103.06495 (2021-03)
Read Like Humans: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Recognition
2103.14146 (2021-03)
Describing and Localizing Multiple Changes with Transformers
2108.00202 (2021-08)
HiFT: Hierarchical Feature Transformer for Aerial Tracking
and 3 more not named here
_get_activation_fn
brick of this paper's assembly: helper
1 other paper
1 later
2412.07236 (2024-12)
CBraMod: A Criss-Cross Brain Foundation Model for EEG Decoding
torch_cast_to_dtype
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
2106.02584 (2021-06)
Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learning

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

Self-supervision is often used for pre-training to foster performance on a downstream task by constructing meaningful representations of samples. Self-supervised learning (SSL) generally involves generating different views of the same sample and thus requires data augmentations that are challenging to construct for tabular data. This constitutes one of the main challenges of self-supervision for structured data. In the present work, we propose a novel augmentation-free SSL method for tabular data. Our approach, T-JEPA, relies on a Joint Embedding Predictive Architecture (JEPA) and is akin to mask reconstruction in the latent space. It involves predicting the latent representation of one subset of features from the latent representation of a different subset within the same sample, thereby learning rich representations without augmentations. We use our method as a pre-training technique and train several deep classifiers on the obtained representation. Our experimental results demonstrate a substantial improvement in both classification and regression tasks, outperforming models trained directly on samples in their original data space. Moreover, T-JEPA enables some methods to consistently outperform or match the performance of traditional methods likes Gradient Boosted Decision Trees. To understand why, we extensively characterize the obtained representations and show that T-JEPA effectively identifies relevant features for downstream tasks without access to the labels. Additionally, we introduce regularization tokens, a novel regularization method critical for training of JEPA-based models on structured data.

For agents

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

get_harvested_code_for_paper("2410.05016")
get_code_for_paper("2410.05016")
have("2410.05016")

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

curl https://syntology.ai/api/ran/2410.05016.json

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

[![Syntology run record](https://syntology.ai/api/ran/2410.05016.svg)](https://syntology.ai/paper/2410.05016)

Connect an agent — have() is free.