SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2601.18546 · 2026

Information Hidden in Gradients of Regression with Target Noise

Katsiaryna Haitsiukevich, Kai Puolamäki, Arash Jamshidi

arXiv · PDF · Open in the Atlas

Code that ran

8 of 8 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; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 6 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/2601.18546.json · badge: /api/ran/2601.18546.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
edahelsinki/noisy_targets canonical (classifier label)
classifier basis: census
8 ran · 0 no recorded run
FunctionStatusWhere it lives
generate_data Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
edahelsinki/noisy_targets/notes/utils.py
code served (permissive licence) · get_code("f30a106451e1487e")
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

ARGS = {
    'n_samples': 100,
    'n_features': 10,
    'scale': 1.0,
    'noise_std': 0.1,
    'w': np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]),
    'b': 0.5
}
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_data_mlp Ran
ran; our contract draft was wrong, not the code
edahelsinki/noisy_targets/notes/utils.py
code served (permissive licence) · get_code("a1ffcf3806c4bf36")
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 torch
import torch.nn as nn
import numpy as np

torch.manual_seed(0)

class SimpleMLPModel(nn.Module):
    def __init__(self, input_dim=4, output_dim=1):
        super().__init__()
        self.linear1 = nn.Linear(input_dim, 8)
        self.relu = nn.ReLU()
        self.linear2 = nn.Linear(8, output_dim)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.relu(x)
        x = self.linear2(x)
        return x

model = SimpleMLPModel(input_dim=4, output_dim=1)

ARGS = {
    "model": model,
    "n_samples": 100,
    "n_features": 4,
    "scale": 1,
    "noise_std": 0.1
}
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 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
get_data_loaders Ran
ran; our contract draft was wrong, not the code
edahelsinki/noisy_targets/notes/utils.py
code served (permissive licence) · get_code("350dd8e43f88e64b")
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, 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.

import numpy as np
import torch

torch.manual_seed(0)

# Create small dummy data
# X: 10 samples, 4 features
# y: 10 samples, 3 targets
X = np.random.randn(10, 4).astype(np.float32)
y = np.random.randn(10, 3).astype(np.float32)

ARGS = {
    "X": X,
    "y": y,
    "batch_size": 2,
    "test_ratio": 0.2,
    "norm_X": False,
    "norm_y": False,
    "return_ystd": False,
}
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 tuple, ndim 1, dtype family float — a tuple 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
get_fmt Ran
ran; our contract draft was wrong, not the code
edahelsinki/noisy_targets/run_utils.py
code served (permissive licence) · get_code("08d5f37e5a45f708")
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 str. 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)

ARGS = {
    'input_value': 0.00005
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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_hidden_layers Ran
ran; our contract draft was wrong, not the code
edahelsinki/noisy_targets/run_mlp_generated.py
code served (permissive licence) · get_code("245196e36577ffa5")
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
import torch.nn as nn

torch.manual_seed(0)

# Based on the function body, get_hidden_layers takes:
# - n_models: an integer representing number of models
# - model_idx: an integer representing a model index

# The function uses n_models in comparisons and list slicing
# The function uses model_idx as a dictionary key
# Valid model_idx values are 0-4 based on get_model_sizes()

ARGS = {
    'n_models': 3,
    'model_idx': 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 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
get_mlp_init Ran
ran; our contract draft was wrong, not the code
edahelsinki/noisy_targets/run_mlp_generated.py
code served (permissive licence) · get_code("ab09f4ccccf40913")
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 MockMLP. 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 as nn
import copy

torch.manual_seed(0)

class MockMLP(nn.Module):
    """Minimal MLP mock that supports the required interface."""
    
    def __init__(self):
        super().__init__()
        self.fc1 = nn.Linear(4, 3)
        self.fc2 = nn.Linear(3, 2)
    
    def get_param_distortion(self):
        """Returns distortion values for parameters."""
        distortions = {}
        for name, param in self.named_parameters():
            distortions[name] = torch.randn_like(param) * 0.01
        return distortions
    
    def set_params(self, w_distortions, distortion_norm_w=3.0, distortion_norm_b=0.1):
        """Sets parameters with distortion scaling."""
        for name, param in self.named_parameters():
            if name in w_distortions:
                distortion = w_distortions[name]
                # Apply distortion based on parameter type
                if 'weight' in name:
                    param.data = param.data + distortion * distortion_norm_w
                elif 'bias' in name:
                    param.data = param.data + distortion * distortion_norm_b

mlp_gen = MockMLP()

ARGS = {
    'mlp_gen': mlp_gen,
    'w_distortion': 3.0,
}
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 MockMLP — a MockMLP output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MockMLP 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_seeds Ran
ran, honoured the contract we drafted
edahelsinki/noisy_targets/run_utils.py
code served (permissive licence) · get_code("398016a9d4e853c8")
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 (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 numpy as np

def get_seeds(n_seeds):
    seeds = np.arange(0, 10 * n_seeds, 10, dtype=int)
    return seeds

ARGS = {
    "n_seeds": 5
}
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 (5,) — shape (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
load_waveenergy_data Ran
ran, honoured the contract we drafted
edahelsinki/noisy_targets/dataset_utils.py
code served (permissive licence) · get_code("028ea687b38aee06")
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 DataFrame. 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 os
import pandas as pd
import tempfile
import zipfile

# Create a temporary directory with test data
temp_dir = tempfile.mkdtemp()

# Create test CSV files
test_data_1 = pd.DataFrame([[1, 2, 3], [4, 5, 6]])
test_data_2 = pd.DataFrame([[7, 8, 9], [10, 11, 12]])

# Test case 1: Create ZIP files
csv_file_1 = os.path.join(temp_dir, "data_1.csv")
csv_file_2 = os.path.join(temp_dir, "data_2.csv")

test_data_1.to_csv(csv_file_1, index=False, header=False)
test_data_2.to_csv(csv_file_2, index=False, header=False)

# Create ZIP files
zip_file_1 = os.path.join(temp_dir, "data_1.csv.zip")
zip_file_2 = os.path.join(temp_dir, "data_2.csv.zip")

with zipfile.ZipFile(zip_file_1, 'w') as zf:
    zf.write(csv_file_1, arcname="data_1.csv")

with zipfile.ZipFile(zip_file_2, 'w') as zf:
    zf.write(csv_file_2, arcname="data_2.csv")

# Clean up the temporary CSV files (keep only ZIPs)
os.remove(csv_file_1)
os.remove(csv_file_2)

ARGS = {
    "folder": temp_dir,
    "is_zip": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family int 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

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 8 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.

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

Second-order information-such as curvature or data covariance-is critical for optimisation, diagnostics, and robustness. However, in many modern settings, only the gradients are observable. We show that the gradients alone can reveal the Hessian, equalling the data covariance Σ for the linear regression. Our key insight is a simple variance calibration: injecting Gaussian noise so that the total target noise variance equals the batch size ensures that the empirical gradient covariance closely approximates the Hessian, even when evaluated far from the optimum. We provide non-asymptotic operator-norm guarantees under sub-Gaussian inputs. We also show that without such calibration, recovery can fail by an Ω(1) factor. The proposed method is practical (a "set target-noise variance to n" rule) and robust (variance O(n) suffices to recover Σ up to scale). Applications include preconditioning for faster optimisation, adversarial risk estimation, and gradient-only training, for example, in distributed systems. We support our theoretical results with experiments on synthetic and real data.

For agents

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

get_harvested_code_for_paper("2601.18546")
get_code_for_paper("2601.18546")
have("2601.18546")

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

curl https://syntology.ai/api/ran/2601.18546.json

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

[![Syntology run record](https://syntology.ai/api/ran/2601.18546.svg)](https://syntology.ai/paper/2601.18546)

Connect an agent — have() is free.