SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.08215 · 2026

Distribution-Free Robust Predict-Then-Optimize in Function Spaces

Ambuj Tewari, Yash Patel

arXiv · PDF · Open in the Atlas

Code that ran

26 of 28 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 3 violated it and 0 have no contract verdict; 23 failures of our instrument; 2 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/2602.08215.json · badge: /api/ran/2602.08215.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
yashpatel5400/fpo canonical (classifier label)
classifier basis: census
26 ran · 2 no recorded run
FunctionStatusWhere it lives
add_hierarchical_offset_2d Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/data.py
code served (permissive licence) · get_code("b0d450e5bb923406")
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 (8, 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

# Create a small 2D array to use as input
f = np.random.randn(8, 6).astype(np.float32)

# Create a random number generator
rng = np.random.default_rng(seed=42)

# Define all arguments
ARGS = {
    "f": f,
    "K_max": 6,
    "sigma": 0.5,
    "rng": rng
}
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 (8, 6) — shape (8, 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
build_legacy_suffix Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/operator_accuracy.py
code served (permissive licence) · get_code("346f3fb157d99d7e")
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
import numpy as np
from collections import namedtuple

torch.manual_seed(0)

class Args:
    def __init__(self, pde_type, grf_alpha, grf_tau, grf_offset_sigma=None, 
                 fiber_core_radius_factor=None, fiber_potential_depth=None,
                 grin_strength=None, evolution_time_T=None, solver_num_steps=None,
                 viscosity_nu=None):
        self.pde_type = pde_type
        self.grf_alpha = grf_alpha
        self.grf_tau = grf_tau
        self.grf_offset_sigma = grf_offset_sigma
        self.fiber_core_radius_factor = fiber_core_radius_factor
        self.fiber_potential_depth = fiber_potential_depth
        self.grin_strength = grin_strength
        self.evolution_time_T = evolution_time_T
        self.solver_num_steps = solver_num_steps
        self.viscosity_nu = viscosity_nu

ARGS = {
    'args': Args(
        pde_type='poisson',
        grf_alpha=1.5,
        grf_tau=0.8,
        grf_offset_sigma=0.3,
        fiber_core_radius_factor=2.5,
        fiber_potential_depth=3.2,
        grin_strength=1.5e-3,
        evolution_time_T=1.0e-2,
        solver_num_steps=100,
        viscosity_nu=2.5e-4
    ),
    'grf_alpha': None
}
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 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
build_suffix Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/collection_sweep.py
code served (permissive licence) · get_code("8baf73eb04f01385")
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
import numpy as np

class ConfigA:
    """Stand-in config class with all required attributes for build_suffix"""
    def __init__(self):
        self.grf_tau = 2.5
        self.grf_offset_sigma = 1.3
        self.fiber_core_radius_factor = 0.8
        self.fiber_potential_depth = 5.2
        self.evolution_time_T = 0.001
        self.solver_num_steps = 100
        self.grin_strength = 1.5e-3
        self.viscosity_nu = 2.1e-4

ARGS = {
    'pde_type': 'poisson',
    'a': ConfigA(),
    'grf_alpha': 1.5
}
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 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
build_suffix Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/operator_accuracy.py
code served (permissive licence) · get_code("ad37cae170cd3854")
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
import numpy as np

class Args:
    def __init__(self, pde_type, grf_alpha, grf_tau, grf_offset_sigma=None, 
                 fiber_core_radius_factor=None, fiber_potential_depth=None,
                 evolution_time_T=None, solver_num_steps=None, grin_strength=None,
                 viscosity_nu=None):
        self.pde_type = pde_type
        self.grf_alpha = grf_alpha
        self.grf_tau = grf_tau
        self.grf_offset_sigma = grf_offset_sigma
        self.fiber_core_radius_factor = fiber_core_radius_factor
        self.fiber_potential_depth = fiber_potential_depth
        self.evolution_time_T = evolution_time_T
        self.solver_num_steps = solver_num_steps
        self.grin_strength = grin_strength
        self.viscosity_nu = viscosity_nu

ARGS = {
    'args': Args(
        pde_type='poisson',
        grf_alpha=1.5,
        grf_tau=0.8,
        grf_offset_sigma=0.3,
        fiber_core_radius_factor=2.0,
        fiber_potential_depth=5.0,
        evolution_time_T=1e-2,
        solver_num_steps=100,
        grin_strength=1.5e-3,
        viscosity_nu=2.5e-4
    ),
    'grf_alpha': None
}
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 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
calculate_I_AB_components_torch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yashpatel5400/fpo/robust_opt.py
code served (permissive licence) · get_code("32542ac2d757d6ae")
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 torch
import numpy as np

torch.manual_seed(0)

# Create small tensors for testing
M_val = 4

# phi_params_torch: shape (M_val - 1,) since the function concatenates [0.0] with it
phi_params_torch = torch.tensor([0.1, 0.2, 0.3], dtype=torch.float64)

# x_s_torch: shape (M_val,) - represents signal values
x_s_torch = torch.tensor([1.0, 0.5, 0.3, 0.2], dtype=torch.float64)

# q_priors_torch: shape (M_val,) - prior probabilities that sum to 1
q_priors_torch = torch.tensor([0.25, 0.25, 0.25, 0.25], dtype=torch.float64)

# eps: small epsilon for numerical stability
eps = 1e-12

ARGS = {
    "phi_params_torch": phi_params_torch,
    "M_val": M_val,
    "x_s_torch": x_s_torch,
    "q_priors_torch": q_priors_torch,
    "eps": eps,
}
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
calculate_conditional_probs_p_j_given_k_torch Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/robust_opt.py
code served (permissive licence) · get_code("f03101ca398d9223")
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)

# M_val is the dimension parameter used throughout
M_val = 4

# phi_full_torch: used as torch tensor, needs .to(device), and passed to torch.cos and torch.sin
# Shape should be (M_val,) based on the loop structure where it's element-wise multiplied with x_s_torch
phi_full_torch = torch.randn(M_val, dtype=torch.float64)

# x_s_torch: used as torch tensor with .device attribute
# Multiplied element-wise with cos/sin of phi_full_torch
# Shape should match phi_full_torch: (M_val,)
x_s_torch = torch.randn(M_val, dtype=torch.float64)

ARGS = {
    "phi_full_torch": phi_full_torch,
    "x_s_torch": x_s_torch,
    "M_val": M_val,
}
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
channels_to_spectrum_complex_torch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yashpatel5400/fpo/model.py
code served (permissive licence) · get_code("5604e74122809770")
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, 3) 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 expects a tensor with shape (2, *, *) where the first dimension
# contains real and imaginary parts. Let's create a small 3D tensor.
channels_mat_real_imag = torch.randn(2, 4, 3)

ARGS = {
    "channels_mat_real_imag": channels_mat_real_imag
}
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 (4, 3) — shape (4, 3) 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
channels_to_spectrum_complex_torch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yashpatel5400/fpo/collection.py
code served (permissive licence) · get_code("821053dc67710bba")
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

torch.manual_seed(0)

# The function expects a tensor of shape [2, K, K] where:
# - First dimension (size 2) holds real and imaginary parts
# - Second and third dimensions (both size K) form a square matrix
# K can be any positive integer; using K=4 for a small tensor

K = 4
channels_mat_real_imag = torch.randn(2, K, K)

ARGS = {
    'channels_mat_real_imag': channels_mat_real_imag
}
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 (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
expected_base Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/collection_sweep.py
code served (permissive licence) · get_code("3d789b6d350a6382")
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 os

ARGS = {
    'results_out': '/tmp/results',
    'pde_type': 1,
    'Nin': 3,
    'Nout': 5,
    'K': 8,
    'rpx': 2.5,
    'alpha_for_radius': 0.75,
    'suffix': 'test_run'
}
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 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
first_existing Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/appendix_collection_viz.py
code served (permissive licence) · get_code("f1767f684708bcc8")
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 os
import tempfile
import shutil

# Create a temporary directory with a test file
temp_dir = tempfile.mkdtemp()
test_file = os.path.join(temp_dir, "test.txt")
with open(test_file, "w") as f:
    f.write("test")

# Define arguments for first_existing
ARGS = {
    "candidates": [
        "/nonexistent/path/one",
        "/nonexistent/path/two",
        test_file,
        "/nonexistent/path/three"
    ],
    "description": "test configuration file"
}
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 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
generate_latex_table Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/robust_opt_sweep.py
code served (permissive licence) · get_code("2c8d997e3ea51631")
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
import numpy as np
from types import SimpleNamespace

torch.manual_seed(0)

# Build results_dict with the expected structure
results_dict = {
    (0.5, 10, 32): {
        'avg_pgm_IAB': 0.7234,
        'avg_nominal_IAB': 0.6891,
        'avg_robust_IAB': 0.7105,
        'p_value_rob_gt_nom': 0.0234,
        'p_value_rob_gt_pgm': 0.3421,
    },
    (0.3, 5, 16): {
        'avg_pgm_IAB': 0.6543,
        'avg_nominal_IAB': 0.7234,
        'avg_robust_IAB': 0.6890,
        'p_value_rob_gt_nom': 0.5678,
        'p_value_rob_gt_pgm': 0.1234,
    },
    (0.7, 15, 64): {
        'avg_pgm_IAB': 0.8123,
        'avg_nominal_IAB': 0.7654,
        'avg_robust_IAB': 0.8456,
        'p_value_rob_gt_nom': 0.0123,
        'p_value_rob_gt_pgm': 0.0456,
    },
}

# Build args object with required attributes
args = SimpleNamespace(
    pde_type='burgers_equation'
)

ARGS = {
    'results_dict': results_dict,
    'args': args,
}
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 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_mode_indices_and_weights Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/calibration.py
code served (permissive licence) · get_code("ba928d4883913b13")
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 = {
    "K_grid_size": 5,
    "d_dimensions": 2,
    "s_coeff": 2.5,
    "nu_coeff": 1.0,
}
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 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
load_and_prepare_dataloaders Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/model.py
code served (permissive licence) · get_code("0f574e35260259c4")
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 torch
import numpy as np
import tempfile
import os

torch.manual_seed(0)

# Create a temporary directory for the dataset file
temp_dir = tempfile.mkdtemp()
dataset_path = os.path.join(temp_dir, 'test_dataset.npz')

# Create synthetic dataset with complex spectra
# Input: full spectra at resolution 4x4
# Target: truncated spectra at resolution 3x3
num_samples = 10
input_res = 4
target_res = 3

gamma_b_train = np.random.randn(num_samples, input_res, input_res) + 1j * np.random.randn(num_samples, input_res, input_res)
gamma_a_snn_target_train = np.random.randn(num_samples, target_res, target_res) + 1j * np.random.randn(num_samples, target_res, target_res)

# Save to npz file
np.savez(dataset_path, gamma_b_train=gamma_b_train, gamma_a_snn_target_train=gamma_a_snn_target_train)

ARGS = {
    'dataset_path': dataset_path,
    'snn_input_res_expected': input_res,
    'snn_target_res_expected': target_res,
    'batch_size': 2,
    'val_split': 0.2,
    'random_seed': 42,
}
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
load_radius Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/appendix_collection_viz.py
code served (permissive licence) · get_code("668c3e16b59ed2f6")
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 os
import numpy as np
import tempfile
import shutil

# Create a temporary directory structure for calibration results
temp_dir = tempfile.mkdtemp()
calib_results_dir = os.path.join(temp_dir, 'calib_results')
os.makedirs(calib_results_dir, exist_ok=True)

# Create calibration data file
pde_type = 1
n_grid_sim_input_ds = 32
snn_output_res = 16
s_theorem = 0.5
nu_theorem = 0.1
suffix = 'test'

calib_filename = f'coverage_data_PDE{pde_type}_thm_s{s_theorem}_nu{nu_theorem}_d2_Nin{n_grid_sim_input_ds}_SNNout{snn_output_res}_NfullThm{n_grid_sim_input_ds}_{suffix}.npz'
calib_path = os.path.join(calib_results_dir, calib_filename)

# Create synthetic calibration data
nominal_coverages = np.array([0.85, 0.90, 0.92, 0.95, 0.97])
quantiles_q_hat_nu = np.array([0.5, 1.0, 1.5, 2.0, 2.5])

np.savez(calib_path, nominal_coverages=nominal_coverages, quantiles_q_hat_nu=quantiles_q_hat_nu)

# Create a minimal args object
class Args:
    def __init__(self):
        self.pde_type = pde_type
        self.n_grid_sim_input_ds = n_grid_sim_input_ds
        self.snn_output_res = snn_output_res
        self.s_theorem = s_theorem
        self.nu_theorem = nu_theorem
        self.calib_results_dir = calib_results_dir
        self.alpha_for_radius = 0.1
        self.collection_radius_scale = 1.5

args_obj = Args()

ARGS = {
    'args': args_obj,
    'suffix': suffix
}
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, 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
pad_to_full_centered Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/collection.py
code served (permissive licence) · get_code("aef3d77c1537f553")
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 (12, 12) 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 centered block (4x4)
block_centered = np.array([
    [1.0, 2.0, 3.0, 4.0],
    [5.0, 6.0, 7.0, 8.0],
    [9.0, 10.0, 11.0, 12.0],
    [13.0, 14.0, 15.0, 16.0]
], dtype=np.float32)

# Full size should be larger (12x12)
K_full = 12

ARGS = {
    "block_centered": block_centered,
    "K_full": K_full
}
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 (12, 12) — shape (12, 12) 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_list Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/operator_accuracy_sweep.py
code served (permissive licence) · get_code("f860fb1d6549a8a1")
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.

import torch
import numpy as np

# Analyze the function: parse_float_list takes one parameter 'raw'
# which can be a list, tuple, or string containing numeric values
# The function converts it to a list of floats

ARGS = {
    'raw': [1.5, 2.3, 3.7, 4.2]
}
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 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
parse_int_list Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/operator_accuracy_sweep.py
code served (permissive licence) · get_code("8dc6aca9ce6ae15c")
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 int; 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 torch
import numpy as np

# parse_int_list takes a single parameter 'raw'
# It handles list/tuple by joining them as strings
# It then processes strings by replacing commas with spaces and splitting
# Finally converts to integers, filtering out empty strings

# Test with a list input
raw_list = [1, 2, 3, 4, 5]

ARGS = {
    'raw': raw_list
}
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 list, ndim 1, dtype family int — 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
poisson_suffix Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/appendix_collection_viz.py
code served (permissive licence) · get_code("fbe8a50d4093bab9")
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 = {
    'grf_alpha': 0.5,
    'grf_tau': 1.2,
    'grf_offset_sigma': 0.3,
}
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
run_script Ran
ran, violated the contract we drafted
yashpatel5400/fpo/calibration_sweep.py
code served (permissive licence) · get_code("56cfef58cb04093a")
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 os
import sys
import tempfile

# Create a temporary Python script to run
script_content = '''
import sys
print("Script executed with args:", sys.argv[1:])
sys.exit(0)
'''

# Create a temporary file for the script
with tempfile.NamedTemporaryFile(mode='w', suffix='.py', delete=False) as f:
    f.write(script_content)
    script_path = f.name

try:
    # Define arguments for run_script
    ARGS = {
        'script_name': script_path,
        'args_list': ['arg1', 'arg2', 42, 3.14],
        'log_prefix': '[TEST] '
    }
finally:
    # Clean up the temporary script after defining ARGS
    # Note: We keep the file since the function will execute it
    pass
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
run_script Ran
ran, violated the contract we drafted
yashpatel5400/fpo/collection_sweep.py
code served (permissive licence) · get_code("b48067ba3bbdeb9d")
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 os
import sys
import tempfile

# Create a minimal dummy script to run
dummy_script_content = """
import sys
print(f"Script executed with args: {sys.argv[1:]}")
sys.exit(0)
"""

# Write the dummy script to a temporary file
with tempfile.NamedTemporaryFile(mode='w', suffix='.py', delete=False) as f:
    f.write(dummy_script_content)
    dummy_script_path = f.name

try:
    # Define arguments for run_script
    ARGS = {
        'script_name': dummy_script_path,
        'args_list': ['arg1', 'arg2', 42, 3.14],
        'log_prefix': '[TEST] ',
        'env_extra': {
            'CUDA_VISIBLE_DEVICES': '0',
            'MY_VAR': 'my_value'
        }
    }
finally:
    # Clean up the temporary script file after defining ARGS
    pass
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
run_script Ran
ran, violated the contract we drafted
yashpatel5400/fpo/robust_opt_sweep.py
code served (permissive licence) · get_code("2996eec91458a373")
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 os
import sys
import tempfile

# Create a temporary directory for our test script
temp_dir = tempfile.mkdtemp()
test_script = os.path.join(temp_dir, 'test_script.py')

# Write a minimal test script that accepts arguments
with open(test_script, 'w') as f:
    f.write("""
import sys
print(f"Script executed with {len(sys.argv) - 1} arguments")
for i, arg in enumerate(sys.argv[1:]):
    print(f"Arg {i}: {arg}")
sys.exit(0)
""")

# Define arguments for run_script
ARGS = {
    'script_name': test_script,
    'args_list': ['arg1', 42, 3.14, 'another_arg'],
    'log_prefix': '[TEST] '
}
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
solve_poisson_2d_rfft Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yashpatel5400/fpo/data.py
code served (permissive licence) · get_code("70df22f5e0184adf")
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 (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 numpy as np

# Create a small complex array in rfft format
# For a 2D real FFT, the shape is (N1, M_rfft) where M_rfft = N2//2 + 1
# Let's use N1=4, N2=6, so M_rfft=4
N1 = 4
M_rfft = 4

# Create a random complex array representing rfft output
np.random.seed(0)
f_hat_r = np.random.randn(N1, M_rfft).astype(np.complex128) + 1j * np.random.randn(N1, M_rfft).astype(np.complex128)

# Make sure the zero frequency component has some value to test enforce_zero_mean
f_hat_r[0, 0] = 5.0 + 3.0j

ARGS = {
    "f_hat_r": f_hat_r,
    "enforce_zero_mean": True
}
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 (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
spectrum_complex_to_channels_torch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yashpatel5400/fpo/model.py
code served (permissive licence) · get_code("c314096fa242334f")
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
import numpy as np

torch.manual_seed(0)

# Create a small complex spectrum matrix [K, K]
K = 4
spectrum_complex = torch.randn(K, K, dtype=torch.complex64)

ARGS = {
    "spectrum_mat_complex": spectrum_complex
}
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 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
spectrum_complex_to_channels_torch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yashpatel5400/fpo/collection.py
code served (permissive licence) · get_code("89729296470c6f30")
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
import numpy as np

torch.manual_seed(0)

# Case 1: Complex tensor input
complex_spectrum = torch.randn(4, 4, dtype=torch.complex64)

ARGS = {
    "spectrum_mat_complex": complex_spectrum
}
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 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
split_step_solver_2d Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yashpatel5400/fpo/data.py
code served (permissive licence) · get_code("cce9a1789dc85313")
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 (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 numpy as np

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

# Define parameters
N = 4  # Grid size (4x4)
dx = 0.5  # Grid spacing
T = 1.0  # Total time
num_steps = 5  # Number of time steps
hbar_val = 1.0545718e-34  # Reduced Planck constant
m_val = 9.10938e-31  # Electron mass

# Create initial wavefunction psi0 as a 2D complex array
psi0 = np.exp(-((np.arange(N) - N//2)**2)[:, None] - ((np.arange(N) - N//2)**2)[None, :]) + 0j

# Create potential grid V_grid as a 2D array
V_grid = np.zeros((N, N), dtype=np.float64)
V_grid[N//2, N//2] = 0.1  # Add a small potential at center

# Construct the ARGS dictionary
ARGS = {
    'V_grid': V_grid,
    'psi0': psi0,
    'N': N,
    'dx': dx,
    'T': T,
    'num_steps': num_steps,
    'hbar_val': hbar_val,
    'm_val': m_val,
}
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 (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
suffix_candidates Ran
ran; our contract draft was wrong, not the code
yashpatel5400/fpo/operator_accuracy.py
code served (permissive licence) · get_code("4e812720bbb00ede")
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.

import torch

class Args:
    def __init__(self, pde_type, grf_alpha=0.5, grf_tau=0.1, grf_offset_sigma=0.05,
                 fiber_core_radius_factor=0.3, fiber_potential_depth=1.5,
                 evolution_time_T=0.01, solver_num_steps=100,
                 grin_strength=1e-2, viscosity_nu=1e-3):
        self.pde_type = pde_type
        self.grf_alpha = grf_alpha
        self.grf_tau = grf_tau
        self.grf_offset_sigma = grf_offset_sigma
        self.fiber_core_radius_factor = fiber_core_radius_factor
        self.fiber_potential_depth = fiber_potential_depth
        self.evolution_time_T = evolution_time_T
        self.solver_num_steps = solver_num_steps
        self.grin_strength = grin_strength
        self.viscosity_nu = viscosity_nu

ARGS = {
    'args': Args(pde_type='poisson'),
    'grf_alpha': 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 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
run_single_calibration_pipeline Not yet run
unverified (harvested, no recorded run)
yashpatel5400/fpo/calibration_sweep.py
code served (permissive licence) · get_code("6f460ddccfb3254c")
How it ran

no invocation recorded.

run_single_robust_opt Not yet run
unverified (harvested, no recorded run)
yashpatel5400/fpo/robust_opt_sweep.py
code served (permissive licence) · get_code("b4feacd478620451")
How it ran

no invocation recorded.

Also present in other papers' code

Of 28 samples lifted from this paper's linked repositories, 28 carry a function identity and 0 do not. Of the 28 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 28 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

The need to rapidly solve PDEs in engineering design workflows has spurred the rise of neural surrogate models. In particular, neural operator models provide a discretization-invariant surrogate by retaining the infinite-dimensional, functional form of their arguments. Despite improved throughput, such methods lack guarantees on accuracy, unlike classical numerical PDE solvers. Optimizing engineering designs under these potentially miscalibrated surrogates thus runs the risk of producing designs that perform poorly upon deployment. In a similar vein, there is growing interest in automated decision-making under black-box predictors in the finite-dimensional setting, where a similar risk of suboptimality exists under poorly calibrated models. For this reason, methods have emerged that produce adversarially robust decisions under uncertainty estimates of the upstream model. One such framework leverages conformal prediction, a distribution-free post-hoc uncertainty quantification method, to provide these estimates due to its natural pairing with black-box predictors. We herein extend this line of conformally robust decision-making to infinite-dimensional function spaces. We first extend the typical conformal prediction guarantees over finite-dimensional spaces to infinite-dimensional Sobolev spaces. We then demonstrate how such uncertainty can be leveraged to robustly formulate engineering design tasks and characterize the suboptimality of the resulting robust optimal designs. We then empirically demonstrate the generality of our functional conformal coverage method across a diverse collection of PDEs, including the Poisson and heat equations, and showcase the significant improvement of such robust design in a quantum state discrimination task.

For agents

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

get_harvested_code_for_paper("2602.08215")
get_code_for_paper("2602.08215")
have("2602.08215")

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

curl https://syntology.ai/api/ran/2602.08215.json

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

[![Syntology run record](https://syntology.ai/api/ran/2602.08215.svg)](https://syntology.ai/paper/2602.08215)

Connect an agent — have() is free.