SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.19131 · ICML · 2026

Test-Time Learning of Causal Structure from Interventional Data

Dongmei Zhang, Wei Chen, Yang Zhang, Qiang Fu, Hkust, Han Shi, Yuxuan Liang, Rui Ding, Bojun Huang

arXiv · PDF · Open in the Atlas

Code that ran

16 of 20 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 5 honoured the contract we drafted, 1 violated it and 1 has no contract verdict; 9 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/2602.19131.json · badge: /api/ran/2602.19131.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
huawei-noah/trustworthyAI canonical (classifier label)
classifier basis: census
13 ran · 2 no recorded run
bvarici/uai2022-intervention-estimation-latents community
Papers with Code archive
2 ran · 1 no recorded run
juangamella/gies community
Papers with Code archive
1 ran · 1 no recorded run
FunctionStatusWhere it lives
DecomposableScore Ran
ran on a synthesized input · constructed an object; no result was computed
juangamella/gies/gies/scores/exp_gauss_int_l0_pen.py
code served (permissive licence) · get_code("2e27bd725c3464b9")
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, DecomposableScore: 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 copy

class SimpleData:
    """Stand-in for data parameter"""
    def __init__(self):
        self.values = {
            0: [1.0, 2.0, 3.0],
            1: [4.0, 5.0, 6.0],
            2: [7.0, 8.0, 9.0]
        }

class SimpleIntervention:
    """Stand-in for intervention parameter"""
    def __init__(self):
        self.interventions = {}

# Create minimal valid arguments
data = SimpleData()
interv = SimpleIntervention()
cache = True
debug = 0

ARGS = {
    "data": data,
    "interv": interv,
    "cache": cache,
    "debug": debug
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DecomposableScore — 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 DecomposableScore 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
build_kernel_bank Ran
ran; our contract draft was wrong, not the code
huawei-noah/trustworthyAI/research/CauKer/CauKer.py
code served (permissive licence) · get_code("8262196db90ae5ce")
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 list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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.

from sklearn.gaussian_process.kernels import ExpSineSquared, DotProduct, RBF, RationalQuadratic, WhiteKernel, ConstantKernel

ARGS = {
    "time_length": 1000
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
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
build_kernel_bank Ran
ran; our contract draft was wrong, not the code
huawei-noah/trustworthyAI/research/CauKer/CauKer_V2.py
code served (permissive licence) · get_code("fc9ef30f935e2adf")
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 list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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.

from sklearn.gaussian_process.kernels import Matern, ExpSineSquared, DotProduct, RBF, RationalQuadratic, WhiteKernel, ConstantKernel

ARGS = {
    "time_length": 168,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
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
clean_state_dict Ran
ran; our contract draft was wrong, not the code
huawei-noah/trustworthyAI/TabbyTSFM/src/tabby/checkpoint.py
code served (permissive licence) · get_code("0e8e78ee9b8acff8")
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 dict. 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)

# Create a state dict with DDP and torch.compile prefixes
state = {
    'module.layer1.weight': torch.randn(4, 3),
    'module.layer1.bias': torch.randn(4),
    '_orig_mod.layer2.weight': torch.randn(5, 4),
    '_orig_mod.layer2.bias': torch.randn(5),
    'module._orig_mod.layer3.weight': torch.randn(2, 3),
    'clean_layer.weight': torch.randn(3, 2),
}

ARGS = {
    'state': state,
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
discover_ucr_datasets Ran
ran; our contract draft was wrong, not the code
huawei-noah/trustworthyAI/TabbyTSFM/benchmarks/classification/ucr_rf_layer_fusion.py
code served (permissive licence) · get_code("d786dca9d2b63a29")
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 dict. 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 tempfile
import shutil
from pathlib import Path

# Create a temporary directory structure that mimics a UCR dataset root
temp_dir = tempfile.mkdtemp()
ucr_root = temp_dir

# Create some dummy UCR dataset files
dataset_names = ['ArrowHead', 'Beef', 'CBF']

for dataset_name in dataset_names:
    # Create TRAIN file
    train_file = Path(ucr_root) / f'{dataset_name}_TRAIN.tsv'
    train_file.write_text('1,2,3,4\n2,3,4,5\n')
    
    # Create corresponding TEST file
    test_file = Path(ucr_root) / f'{dataset_name}_TEST.tsv'
    test_file.write_text('3,4,5,6\n4,5,6,7\n')

# Also create an adjusted version directory for one dataset
adjusted_dir = Path(ucr_root) / 'Missing_value_and_variable_length_datasets_adjusted'
adjusted_dir.mkdir(exist_ok=True)

adjusted_dataset = 'ArrowHead'
adjusted_train = adjusted_dir / f'{adjusted_dataset}_TRAIN.tsv'
adjusted_train.write_text('1,2,3,4\n2,3,4,5\n')

adjusted_test = adjusted_dir / f'{adjusted_dataset}_TEST.tsv'
adjusted_test.write_text('3,4,5,6\n4,5,6,7\n')

ARGS = {
    'ucr_root': ucr_root
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
gen_data_given_model Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
huawei-noah/trustworthyAI/datasets/synthetic_datasets.py
code served (permissive licence) · get_code("89e14b6833a1e17a")
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 numpy as np
from copy import deepcopy

np.random.seed(0)

# Create a strictly lower triangular coefficient matrix
# n_features = 4
n_features = 4
b = np.tril(np.random.randn(n_features, n_features), k=-1)

# Create scales for disturbance variables
s = np.abs(np.random.randn(n_features)) + 0.5

# Create means of observed variables
c = np.random.randn(n_features)

ARGS = {
    'b': b,
    's': s,
    'c': c,
    'n_samples': 10,
    'noise_type': 'lingam',
    'permutate': False,
}
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 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
generate_W Ran
ran, honoured the contract we drafted
huawei-noah/trustworthyAI/datasets/synthetic_datasets.py
code served (permissive licence) · get_code("26ea7751b1dcda50")
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 2, dtype family float, shape (6, 6) 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

ARGS = {
    "d": 6,
    "prob": 0.5,
    "low": 0.5,
    "high": 2.0,
}
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 2, dtype family float, shape (6, 6) — shape (6, 6) 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_session_maker Ran
ran; our contract draft was wrong, not the code
huawei-noah/trustworthyAI/gcastle/web/models/models.py
code served (permissive licence) · get_code("5a857da4a9d09551")
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 sessionmaker. 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.

from sqlalchemy import create_engine

# Create an in-memory SQLite database engine
engine = create_engine('sqlite:///:memory:')

ARGS = {
    'engine': engine
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
sqlalchemy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type sessionmaker — a sessionmaker output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a sessionmaker 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
is_hf_snapshot Ran
ran, violated the contract we drafted
huawei-noah/trustworthyAI/TabbyTSFM/src/tabby/checkpoint.py
code served (permissive licence) · get_code("f64090e43dff9b46")
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 tempfile
import os
from pathlib import Path

# Create a temporary directory structure that mimics a HuggingFace snapshot
temp_dir = tempfile.mkdtemp()
config_path = Path(temp_dir) / 'config.json'
weights_path = Path(temp_dir) / 'model.safetensors'

# Create the required files
config_path.write_text('{"model_type": "test"}')
weights_path.write_text('dummy weights')

ARGS = {
    'path': temp_dir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — 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
model_quantile_levels Ran
ran, honoured the contract we drafted
huawei-noah/trustworthyAI/TabbyTSFM/src/tabby/predictor.py
code served (permissive licence) · get_code("103568c0fa4e1ea4")
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 float, shape (99,) 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

ARGS = {
    "num_quantiles": 99
}
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 float, shape (99,) — shape (99,) 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
parse_float Ran
ran, honoured the contract we drafted
huawei-noah/trustworthyAI/TabbyTSFM/benchmarks/classification/ucr_rf_layer_fusion.py
code served (permissive licence) · get_code("1a5e89a8e5f54553")
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 float, 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 numpy as np

def parse_float(token: str) -> float:
    """Parse UCR values while treating common missing tokens as NaN."""
    token = token.strip()
    if token == '' or token.lower() in {'nan', 'na', 'null', 'none', '?'}:
        return float('nan')
    return float(token)

ARGS = {
    'token': '42.5'
}
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 float, 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
post_pasp Ran
ran; our contract draft was wrong, not the code
bvarici/uai2022-intervention-estimation-latents/main_codes/functions_sample.py
pointer only (licence: NONE) · get_code("600e9c635690e64c")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, bvarici/uai2022-intervention-estimation-latents 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.

random_binary_map Ran
ran, honoured the contract we drafted
huawei-noah/trustworthyAI/research/CauKer/CauKer.py
code served (permissive licence) · get_code("9d2ab88ca6670a55")
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 2, dtype family float, shape (2, 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

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

# Create two small numpy arrays to serve as kernels
a = np.array([[1.0, 2.0], [3.0, 4.0]])
b = np.array([[5.0, 6.0], [7.0, 8.0]])

ARGS = {
    "a": a,
    "b": b
}
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 2, dtype family float, shape (2, 2) — shape (2, 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
remove_supersets Ran
ran; our contract draft was wrong, not the code
bvarici/uai2022-intervention-estimation-latents/main_codes/functions_sample.py
pointer only (licence: NONE) · get_code("eca6afdd5d1dd322")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, bvarici/uai2022-intervention-estimation-latents 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.

resolve_checkpoint_path Ran
ran; our contract draft was wrong, not the code
huawei-noah/trustworthyAI/TabbyTSFM/src/tabby/checkpoint.py
code served (permissive licence) · get_code("715bc8b016f730f7")
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 PosixPath. 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 tempfile
import json
from pathlib import Path

# Create a temporary directory structure for testing
temp_dir = tempfile.mkdtemp()
temp_path = Path(temp_dir)

# Test case 1: Create a valid HF snapshot directory
hf_snapshot_dir = temp_path / "hf_snapshot"
hf_snapshot_dir.mkdir(exist_ok=True)
(hf_snapshot_dir / "config.json").write_text('{"model_type": "test"}')
(hf_snapshot_dir / "model.safetensors").write_text("fake_weights")

# Test case 2: Create a step directory with pytorch_model.bin
step_dir = temp_path / "checkpoints" / "step_1000"
step_dir.mkdir(parents=True, exist_ok=True)
(step_dir / "pytorch_model.bin").write_text("fake_pytorch_weights")

# Test case 3: Create a directory with latest symlink
checkpoint_parent = temp_path / "checkpoint_parent"
checkpoint_parent.mkdir(exist_ok=True)
latest_target = checkpoint_parent / "latest_checkpoint"
latest_target.mkdir(exist_ok=True)
(latest_target / "config.json").write_text('{"model_type": "test"}')
(latest_target / "model.safetensors").write_text("fake_weights")
latest_link = checkpoint_parent / "latest"
latest_link.symlink_to(latest_target)

# Use the HF snapshot directory as the checkpoint
checkpoint = str(hf_snapshot_dir)

ARGS = {
    "checkpoint": checkpoint
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
huggingface_hub — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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
zero_mean Ran
ran, honoured the contract we drafted
huawei-noah/trustworthyAI/research/CauKer/CauKer.py
code served (permissive licence) · get_code("b40263c4308a6bb7")
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 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 numpy as np

# Create a small numpy array for testing
x = np.array([[1.0, 2.0, 3.0, 4.0],
              [5.0, 6.0, 7.0, 8.0]])

ARGS = {
    "x": x
}
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 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
ExpGaussIntL0Pen Not yet run
unverified (harvested, no recorded run)
juangamella/gies/gies/scores/exp_gauss_int_l0_pen.py
code served (permissive licence) · get_code("ffbcd63750cbe3c8")
How it ran

no invocation recorded.

IMAG_pasp_sample Not yet run
unverified (harvested, no recorded run)
bvarici/uai2022-intervention-estimation-latents/main_codes/functions_sample.py
pointer only (licence: NONE) · get_code("2e08c59089a257b4")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, bvarici/uai2022-intervention-estimation-latents 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. No invocation recorded.

gen_data_given_model_2nd_order Not yet run
unverified (harvested, no recorded run)
huawei-noah/trustworthyAI/datasets/synthetic_datasets.py
code served (permissive licence) · get_code("44b8e3c0e4f39a19")
How it ran

no invocation recorded.

transform Not yet run
unverified (harvested, no recorded run)
huawei-noah/trustworthyAI/competition/PCIC2021/submission.py
code served (permissive licence) · get_code("a139780545cc072f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 20 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 16 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
build_kernel_bank 1 other paper
1 dated earlier
2508.02879 (2025-08)
CauKer: Classification Time Series Foundation Models Can Be Pretrained on Synthetic Data
generate_W 1 other paper
1 dated earlier
2106.01474 (2021-06)
Testing Directed Acyclic Graph via Structural, Supervised and Generative Adversarial Learning
random_binary_map 1 other paper
1 dated earlier
2508.02879 (2025-08)
CauKer: Classification Time Series Foundation Models Can Be Pretrained on Synthetic Data
zero_mean 1 other paper
1 dated earlier
2508.02879 (2025-08)
CauKer: Classification Time Series Foundation Models Can Be Pretrained on Synthetic Data

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

Supervised causal learning has shown promise in causal discovery, yet it often struggles with generalization across diverse interventional settings, particularly when intervention targets are unknown. To address this, we propose TICL (Test-time Interventional Causal Learning), a novel method that synergizes Test-Time Training with Joint Causal Inference. Specifically, we design a self-augmentation strategy to generate instance-specific training data at test time, effectively avoiding distribution shifts. Furthermore, by integrating joint causal inference, we developed a PC-inspired two-phase supervised learning scheme, which effectively leverages self-augmented training data while ensuring theoretical identifiability. Extensive experiments on bnlearn benchmarks demonstrate TICL's superiority in multiple aspects of causal discovery and intervention target detection.

For agents

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

get_harvested_code_for_paper("2602.19131")
get_code_for_paper("2602.19131")
have("2602.19131")

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

curl https://syntology.ai/api/ran/2602.19131.json

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

[![Syntology run record](https://syntology.ai/api/ran/2602.19131.svg)](https://syntology.ai/paper/2602.19131)

Connect an agent — have() is free.