SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1908.10507 · 2019

Forward-Mode Differentiation of Maxwell's Equations

arXiv · PDF · Open in the Atlas

Code that ran

14 of 21 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, 0 violated it and 14 have no contract verdict; 0 failures of our instrument; 7 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; 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/1908.10507.json · badge: /api/ran/1908.10507.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
twhughes/ceviche canonical (classifier label)
classifier basis: census
14 ran · 7 no recorded run
FunctionStatusWhere it lives
abs Ran
ran on a synthesized input
twhughes/ceviche/ceviche/viz.py
code served (permissive licence) · get_code("480a78607135a128")
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 Axes. 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
import matplotlib.pyplot as plt

# Create a small 2D array for val
val = np.random.randn(4, 3)

# Create a small 2D array for outline
outline = np.random.randn(4, 3)

# Create a matplotlib axes object
fig, ax = plt.subplots(1, 1, constrained_layout=True)

ARGS = {
    'val': val,
    'outline': outline,
    'ax': ax,
    'cbar': False,
    'cmap': 'magma',
    'outline_alpha': 0.5,
    'outline_val': None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Axes — a Axes output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Axes 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
adam_optimize Ran
ran on a synthesized input
twhughes/ceviche/ceviche/optimizers.py
code served (permissive licence) · get_code("195bf3af50815188")
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

def simple_objective(params):
    """Simple quadratic objective: sum of squares"""
    return np.sum(params ** 2)

def simple_jac(params):
    """Gradient of sum of squares: 2 * params"""
    return 2 * params

def dummy_callback(iteration, of_list, params):
    """Dummy callback that does nothing"""
    pass

ARGS = {
    'objective': simple_objective,
    'params': np.array([1.5, -2.3, 0.8, -1.2]),
    'jac': simple_jac,
    'step_size': 0.01,
    'Nsteps': 5,
    'bounds': None,
    'direction': 'min',
    'beta1': 0.9,
    'beta2': 0.999,
    'callback': dummy_callback,
    'verbose': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
autograd, 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
compute_f Ran
ran on a synthesized input
twhughes/ceviche/ceviche/sources.py
code served (permissive licence) · get_code("175fcc923c54404c")
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 (20,) 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 autograd.numpy as npa
import numpy as np

# Analyze function signature:
# - theta: angle parameter (scalar or array)
# - lambda0: wavelength parameter (scalar)
# - dL: spacing parameter (scalar)
# - shape: tuple for array shape

# Construct valid arguments
theta = 0.3  # scalar angle in radians
lambda0 = 1.0  # scalar wavelength
dL = 0.1  # scalar spacing
shape = (4, 5)  # distinct dimensions for probing

ARGS = {
    "theta": theta,
    "lambda0": lambda0,
    "dL": dL,
    "shape": shape,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
autograd, 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 (20,) — shape (20,) 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
curl_E Ran
ran on a synthesized input
twhughes/ceviche/ceviche/derivatives.py
code served (permissive licence) · get_code("e1b020dc610f81eb")
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 3, dtype family float, shape (2, 3, 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 autograd.numpy as npa

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

# Create small 3D arrays for the electric field components
# Using distinct dimensions: (2, 3, 4) for batch, y-axis, z-axis
batch_size = 2
dim_y = 3
dim_z = 4

Ex = npa.random.randn(batch_size, dim_y, dim_z)
Ey = npa.random.randn(batch_size, dim_y, dim_z)
Ez = npa.random.randn(batch_size, dim_y, dim_z)

# Spatial grid spacing
dL = 0.1

# Axis parameter (0, 1, or 2)
axis = 0

ARGS = {
    "axis": axis,
    "Ex": Ex,
    "Ey": Ey,
    "Ez": Ez,
    "dL": dL,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
autograd — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 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
curl_H Ran
ran on a synthesized input
twhughes/ceviche/ceviche/derivatives.py
code served (permissive licence) · get_code("c0c5035e89b60e25")
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 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 autograd.numpy as npa
import numpy as np

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

# Create small 3D arrays for the magnetic field components
# Shape: (2, 4, 4) - batch=2, spatial dims=4x4
Hx = npa.random.randn(2, 4, 4)
Hy = npa.random.randn(2, 4, 4)
Hz = npa.random.randn(2, 4, 4)

# Grid spacing parameter
dL = 0.1

# Axis parameter (can be 0, 1, or 2)
axis = 0

ARGS = {
    "axis": axis,
    "Hx": Hx,
    "Hy": Hy,
    "Hz": Hz,
    "dL": dL,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
autograd, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, 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
get_entries_indices Ran
ran on a synthesized input
twhughes/ceviche/ceviche/utils.py
code served (permissive licence) · get_code("8b0444e741bdb47d")
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 autograd.numpy as npa
import scipy.sparse as sp
import numpy as np

# Create a small CSR matrix for testing
# 4x3 matrix with some non-zero entries
data = np.array([1.0, 2.0, 3.0, 4.0])
row = np.array([0, 1, 2, 3])
col = np.array([0, 1, 2, 0])
csr_matrix = sp.csr_matrix((data, (row, col)), shape=(4, 3))

ARGS = {
    "csr_matrix": csr_matrix
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
autograd, numpy, scipy — 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
grad_sp_mult_entries_reverse Ran
ran on a synthesized input
twhughes/ceviche/ceviche/primitives.py
code served (permissive licence) · get_code("32efd6e78703c636")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned function. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyzing the function:
# - ans: appears to be unused in the function body
# - entries: unused in the function body
# - indices: tuple (ia, ja) of index arrays/tensors used to index into ans and x
# - x: tensor that gets indexed by ja
# The vjp function returns v[ia] * x[ja], so:
#   - v should be indexable (tensor or array)
#   - ia should be valid indices for v
#   - ja should be valid indices for x
#   - x should support element-wise multiplication

# Create small tensors
ans_tensor = torch.randn(2, 4)  # arbitrary, unused in function
entries_tensor = torch.randn(3)  # arbitrary, unused in function
x_tensor = torch.randn(4)  # will be indexed by ja

# Create index arrays
# ia will index into v (which has same shape as ans output)
ia = torch.tensor([0, 1, 0], dtype=torch.long)  # 3 indices, values < 2
ja = torch.tensor([0, 1, 2], dtype=torch.long)  # 3 indices, values < 4

indices = (ia, ja)

ARGS = {
    "ans": ans_tensor,
    "entries": entries_tensor,
    "indices": indices,
    "x": x_tensor,
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
jacobian Ran
ran on a synthesized input
twhughes/ceviche/ceviche/jacobians.py
code served (permissive licence) · get_code("f1957dddb0eee8e5")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned function. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import autograd.numpy as npa
from autograd.core import make_vjp, make_jvp
from autograd.wrap_util import unary_to_nary
from autograd.extend import vspace

def simple_function(x):
    """A simple test function: f(x) = x^2 + 2*x"""
    return npa.sum(x ** 2 + 2 * x)

def vector_function(x):
    """A function that returns a vector output"""
    return npa.array([x[0] ** 2, x[1] * 2, x[0] + x[1]])

# Test input - small 1D array
test_input = npa.array([1.0, 2.0, 3.0, 4.0])

ARGS = {
    'fun': simple_function,
    'argnum': 0,
    'mode': 'reverse',
    'step_size': 1e-06
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
autograd — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
jacobian_forward Ran
ran on a synthesized input
twhughes/ceviche/ceviche/jacobians.py
code served (permissive licence) · get_code("6fa0121267278eeb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned function. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import autograd.numpy as npa
from autograd.core import make_vjp, make_jvp
from autograd.wrap_util import unary_to_nary
from autograd.extend import vspace

def float_2_array(x):
    """Convert float-like to array"""
    if isinstance(x, (int, float, complex)):
        return npa.array([x])
    return npa.asarray(x)

def simple_function(x):
    """Simple test function: f(x) = x^2 + 2*x"""
    return x**2 + 2*x

def simple_function_vector(x):
    """Vector function: f(x) = [x[0]^2, x[1]*x[0], x[2]+x[3]]"""
    return npa.array([x[0]**2, x[1]*x[0], x[2] + x[3]])

# Test with a simple scalar function
x_scalar = npa.array(3.0)

# Test with a vector input
x_vector = npa.array([1.0, 2.0, 3.0, 4.0])

ARGS = {
    "fun": simple_function_vector,
    "x": x_vector
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
autograd — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
jacobian_reverse Ran
ran on a synthesized input
twhughes/ceviche/ceviche/jacobians.py
code served (permissive licence) · get_code("11d243fbf6c19486")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned function. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import autograd.numpy as npa
from autograd.core import make_vjp, make_jvp
from autograd.wrap_util import unary_to_nary
from autograd.extend import vspace
import numpy as np

def float_2_array(x):
    """Convert float or array to numpy array"""
    if isinstance(x, (int, float)):
        return npa.array([x])
    return npa.asarray(x)

def simple_fun(x):
    """Simple test function: x^2 + 2*x"""
    return x**2 + 2*x

def vector_fun(x):
    """Vector-valued function: [x[0]^2, x[1]*x[0], x[1]^2]"""
    return npa.array([x[0]**2, x[1]*x[0], x[1]**2])

def matrix_fun(x):
    """Matrix-valued function that takes a vector and outputs a matrix"""
    # Input: vector of shape (4,)
    # Output: matrix of shape (3, 2)
    x_flat = x.flatten()
    return npa.array([
        [x_flat[0]**2, x_flat[1]*x_flat[2]],
        [x_flat[2] + x_flat[3], x_flat[0]*x_flat[1]],
        [x_flat[3]**2, x_flat[0] + x_flat[1]]
    ])

# Test inputs
scalar_input = 2.5
vector_input = npa.array([1.0, 2.0, 3.0, 4.0])
matrix_input = npa.array([1.0, 2.0, 3.0, 4.0])

ARGS = {
    "fun": vector_fun,
    "x": vector_input
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
autograd, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
real Ran
ran on a synthesized input
twhughes/ceviche/ceviche/viz.py
code served (permissive licence) · get_code("f559c4723f95f006")
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 Axes. 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
import matplotlib.pyplot as plt

# Create a small 2D array for the main value
val = np.array([[1.0 + 0.5j, 2.0 - 0.3j, -1.5 + 0.2j, 0.8 - 0.1j],
                 [0.5 + 1.2j, -0.7 + 0.9j, 1.1 - 0.4j, -0.3 + 0.6j],
                 [2.0 - 0.5j, -1.0 + 1.5j, 0.2 + 0.8j, 1.3 - 0.9j]], dtype=complex)

# Create a small 2D array for the outline
outline = np.array([[0.1, 0.2, 0.15, 0.05],
                    [0.3, 0.4, 0.35, 0.25],
                    [0.2, 0.25, 0.2, 0.15]], dtype=float)

# Create a matplotlib axes object
fig, ax = plt.subplots(1, 1, constrained_layout=True)

ARGS = {
    'val': val,
    'outline': outline,
    'ax': ax,
    'cbar': True,
    'cmap': 'RdBu',
    'outline_alpha': 0.5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Axes — a Axes output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Axes 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
solver_eigs Ran
ran on a synthesized input
twhughes/ceviche/ceviche/modes.py
code served (permissive licence) · get_code("90fb0f26be8ac45b")
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 numpy as np
import scipy.sparse as sp
import scipy.sparse.linalg as spl

torch.manual_seed(0)

# Create a small sparse matrix as a linear operator
# The function expects A to work with scipy.sparse.linalg.eigs
# eigs requires a square matrix or LinearOperator

# Create a small 4x4 sparse symmetric matrix (needed for eigs to work properly)
data = np.array([2.0, 3.0, 1.0, 4.0, 1.5, 2.5])
row = np.array([0, 0, 1, 1, 2, 3])
col = np.array([0, 1, 1, 2, 2, 3])
A_sparse = sp.csr_matrix((data, (row, col)), shape=(4, 4))

# Make it symmetric for eigs to work well
A_sparse = A_sparse + A_sparse.T

# Convert to a LinearOperator which is what eigs expects
A = sp.linalg.aslinearoperator(A_sparse)

# Number of eigenmodes to compute (must be less than matrix size)
Neigs = 2

# Estimate for eigenvalues (used as sigma parameter in eigs)
guess_value = 1.0

ARGS = {
    "A": A,
    "Neigs": Neigs,
    "guess_value": guess_value
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, 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
step_adam Ran
ran on a synthesized input
twhughes/ceviche/ceviche/optimizers.py
code served (permissive licence) · get_code("e77506199877251d")
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 2, 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

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

# Create small gradient array (shape 4,)
gradient = np.array([0.1, -0.2, 0.3, -0.15], dtype=np.float32)

# Create momentum term (same shape as gradient)
mopt_old = np.array([0.05, -0.1, 0.15, -0.08], dtype=np.float32)

# Create velocity term (same shape as gradient)
vopt_old = np.array([0.02, 0.03, 0.01, 0.04], dtype=np.float32)

# Iteration counter
iteration = 5

# Beta1 (exponential decay rate for first moment)
beta1 = 0.9

# Beta2 (exponential decay rate for second moment)
beta2 = 0.999

# Epsilon for numerical stability
epsilon = 1e-08

ARGS = {
    "gradient": gradient,
    "mopt_old": mopt_old,
    "vopt_old": vopt_old,
    "iteration": iteration,
    "beta1": beta1,
    "beta2": beta2,
    "epsilon": epsilon
}
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 2, 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
transpose_indices Ran
ran on a synthesized input
twhughes/ceviche/ceviche/utils.py
code served (permissive licence) · get_code("61e877158b641cc2")
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 int, shape (3, 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 autograd.numpy as npa
import numpy as np

def transpose_indices(indices):
    return npa.flip(indices, axis=0)

# Create a small numpy array to serve as indices
# Using a 2D array with shape (3, 4) for clarity
indices = np.array([
    [0, 1, 2, 3],
    [4, 5, 6, 7],
    [8, 9, 10, 11]
], dtype=np.int32)

ARGS = {
    "indices": indices
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
autograd, 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 int, shape (3, 4) — shape (3, 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
b_TFSF Not yet run
unverified (harvested, no recorded run)
twhughes/ceviche/ceviche/sources.py
code served (permissive licence) · get_code("48d1890792d089d5")
How it ran

no invocation recorded.

compute_Q Not yet run
unverified (harvested, no recorded run)
twhughes/ceviche/ceviche/sources.py
code served (permissive licence) · get_code("3515683a16b35b70")
How it ran

no invocation recorded.

compute_derivative_matrices Not yet run
unverified (harvested, no recorded run)
twhughes/ceviche/ceviche/derivatives.py
code served (permissive licence) · get_code("fc7f8ba239f1a047")
How it ran

no invocation recorded.

grad_sp_mult_x_reverse Not yet run
unverified (harvested, no recorded run)
twhughes/ceviche/ceviche/primitives.py
code served (permissive licence) · get_code("02bb35d49602cd88")
How it ran

no invocation recorded.

make_sparse Not yet run
unverified (harvested, no recorded run)
twhughes/ceviche/ceviche/utils.py
code served (permissive licence) · get_code("e3020417beb6b8ca")
How it ran

no invocation recorded.

solve_linear Not yet run
unverified (harvested, no recorded run)
twhughes/ceviche/ceviche/solvers.py
code served (permissive licence) · get_code("ae7d57329e86bec6")
How it ran

no invocation recorded.

sp_mult Not yet run
unverified (harvested, no recorded run)
twhughes/ceviche/ceviche/primitives.py
code served (permissive licence) · get_code("149de24946952497")
How it ran

no invocation recorded.

Also present in other papers' code

Of 21 samples lifted from this paper's linked repositories, 21 carry a function identity and 0 do not. Of the 21 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 21 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

We present a previously unexplored forward-mode differentiation method for Maxwell's equations, with applications in the field of sensitivity analysis. This approach yields exact gradients and is similar to the popular adjoint variable method, but provides a significant improvement in both memory and speed scaling for problems involving several output parameters, as we analyze in the context of finite-difference time-domain (FDTD) simulations. Furthermore, it provides an exact alternative to numerical derivative methods, based on finite-difference approximations. To demonstrate the usefulness of the method, we perform sensitivity analysis of two problems. First we compute how the spatial near-field intensity distribution of a scatterer changes with respect to its dielectric constant. Then, we compute how the spectral power and coupling efficiency of a surface grating coupler changes with respect to its fill factor.

For agents

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

get_harvested_code_for_paper("1908.10507")
get_code_for_paper("1908.10507")
have("1908.10507")

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

curl https://syntology.ai/api/ran/1908.10507.json

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

[![Syntology run record](https://syntology.ai/api/ran/1908.10507.svg)](https://syntology.ai/paper/1908.10507)

Connect an agent — have() is free.