SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.02645 · ICLR · 2024

Astral: training physics-informed neural networks with error majorants

Vladimir Fanaskov, Tianchi Yu, Alexander Rudikov

arXiv · PDF · Open in the Atlas

Code that ran

11 of 13 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, 2 violated it and 0 have no contract verdict; 7 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/2406.02645.json · badge: /api/ran/2406.02645.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
4gnskq5g2s-collab/Astral canonical (classifier label)
classifier basis: census
11 ran · 2 no recorded run
FunctionStatusWhere it lives
draw_coefficients Ran
ran; our contract draft was wrong, not the code
4gnskq5g2s-collab/Astral/datasets/Convection_Diffusion.py
code served (permissive licence) · get_code("6c0464c49f680fda")
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 ArrayImpl, ndim 3, dtype family float, shape (2, 1, 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 jax
import jax.numpy as jnp
from jax import random

# Set a seed for reproducibility
key = random.PRNGKey(0)

# Create a small weights array
# The function uses weights.shape, so we need a real array
weights = jnp.array([[1.0, 2.0, 3.0, 4.0],
                      [5.0, 6.0, 7.0, 8.0]])

ARGS = {
    "key": key,
    "weights": weights
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 3, dtype family float, shape (2, 1, 4) — shape (2, 1, 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
gamma Ran
ran, violated the contract we drafted
4gnskq5g2s-collab/Astral/datasets/elastoplasticity.py
code served (permissive licence) · get_code("bc6fc65a144102d4")
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 ArrayImpl, ndim 1, dtype family float, shape (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 jax.numpy as jnp
import numpy as np

# The function gamma takes 4 parameters: u, mu, k_star, delta
# All are used in numerical operations with jnp functions
# u is compared with t0 and used in division
# mu, k_star, delta are used in arithmetic operations

# Create small scalar or array values
u_val = jnp.array([0.5, 2.0, 3.5, 1.0])  # array of test values
mu_val = jnp.array(1.5)  # scalar
k_star_val = jnp.array(2.0)  # scalar
delta_val = jnp.array(0.5)  # scalar

ARGS = {
    "u": u_val,
    "mu": mu_val,
    "k_star": k_star_val,
    "delta": delta_val,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 1, dtype family float, shape (4,) — shape (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_basis Ran
ran, violated the contract we drafted
4gnskq5g2s-collab/Astral/datasets/L_shaped.py
code served (permissive licence) · get_code("276bb5432f460935")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (2, 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 numpy as np

# Analyze the function:
# - f: used in a list comprehension, iterated over with "for f_ in f"
#      each f_ is expanded and multiplied with x
#      f_ should be array-like that can be multiplied with x
# - x: used in np.expand_dims(f_, 0) * x and np.sum(..., axis=1)
#      should be a 2D array where axis=1 makes sense
#      x should have shape (n_samples, n_features)

# Set up small arrays:
# f: frequency basis vectors, shape (n_freqs, n_features)
# x: input data, shape (n_samples, n_features)

np.random.seed(0)

f = np.array([
    [1.0, 2.0, 0.5, 1.5],
    [0.3, 0.7, 1.2, 0.9],
    [2.1, 0.1, 1.8, 0.4]
], dtype=np.complex128)

x = np.array([
    [0.1, 0.2, 0.3, 0.4],
    [0.5, 0.6, 0.7, 0.8]
], dtype=np.float64)

ARGS = {
    "f": f,
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (2, 3) — shape (2, 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
get_coefficients Ran
ran; our contract draft was wrong, not the code
4gnskq5g2s-collab/Astral/datasets/L_shaped.py
code served (permissive licence) · get_code("d3a0f8d6286b9f45")
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 numpy as np
import jax.numpy as jnp
from jax import random

# Create a JAX random key
key = random.PRNGKey(0)

# Create a small tensor for w
# The function uses w.shape, so w needs to be array-like with a shape attribute
w = np.zeros((2, 3, 4), dtype=np.complex64)

ARGS = {
    "key": key,
    "w": w,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, 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, 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
get_dataset Ran
ran; our contract draft was wrong, not the code
4gnskq5g2s-collab/Astral/datasets/magnetostatics.py
code served (permissive licence) · get_code("3d70b75be6df62eb")
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 jax
import jax.numpy as jnp
from jax import random

# Initialize random key
key = random.PRNGKey(0)

# Number of samples to generate
N_samples = 2

ARGS = {
    "key": key,
    "N_samples": N_samples,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, scipy, sympy — 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_eigvalues Ran
ran, honoured the contract we drafted
4gnskq5g2s-collab/Astral/datasets/Convection_Diffusion.py
code served (permissive licence) · get_code("77d4bfafd1b02272")
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 ArrayImpl, ndim 2, dtype family float, shape (4, 1) 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 jax.numpy as jnp

# Analyze the function:
# - N: used as input to jnp.arange(N), should be an integer
# - a: used in arithmetic operations (a ** 2 / 4), should be a scalar or array

# The function computes eigenvalues using N and parameter a
# N should be a small integer for tiny tensors
# a should be a scalar value

N = 4
a = 2.0

ARGS = {
    "N": N,
    "a": a,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 2, dtype family float, shape (4, 1) — shape (4, 1) 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_sample Ran
ran; our contract draft was wrong, not the code
4gnskq5g2s-collab/Astral/datasets/magnetostatics.py
code served (permissive licence) · get_code("476c537d47b2e48b")
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 jax
import jax.numpy as jnp
from jax import random

# Set seed for reproducibility
key = random.PRNGKey(0)

# Create small coordinate arrays
# coords_train: shape (N_train, 2) - training coordinates
N_train = 4
coords_train = jnp.array([
    [0.1, 0.2],
    [0.3, 0.4],
    [0.5, 0.6],
    [0.7, 0.8]
])

# coords_legendre: shape (N_legendre, 2) - Legendre quadrature coordinates
N_legendre = 3
coords_legendre = jnp.array([
    [0.15, 0.25],
    [0.35, 0.45],
    [0.55, 0.65]
])

# coords_eval: shape (N_eval, 2) - evaluation coordinates
N_eval = 2
coords_eval = jnp.array([
    [0.2, 0.3],
    [0.6, 0.7]
])

# weights_: dummy weights parameter (appears in signature but not used in body)
weights_ = jnp.array([1.0, 1.0, 1.0, 1.0])

ARGS = {
    'key': key,
    'coords_train': coords_train,
    'weights_': weights_,
    'coords_legendre': coords_legendre,
    'coords_eval': coords_eval
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, sympy — 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_weights Ran
ran, honoured the contract we drafted
4gnskq5g2s-collab/Astral/datasets/Convection_Diffusion.py
code served (permissive licence) · get_code("afa8720aaf616f2c")
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 ArrayImpl, ndim 1, dtype family float, shape (10,) 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 jax.numpy as jnp

# Analyze the function:
# - N: integer, used as upper limit in arange
# - L: numeric value, used in division
# - s: numeric value, used as exponent
# - p: numeric value, used as exponent

# All parameters are used in arithmetic operations with jnp functions
# They should be simple numeric values (int or float)

ARGS = {
    "N": 10,
    "L": 5.0,
    "s": 2.0,
    "p": 1.5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 1, dtype family float, shape (10,) — shape (10,) 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_weights_and_frequencies Ran
ran; our contract draft was wrong, not the code
4gnskq5g2s-collab/Astral/datasets/L_shaped.py
code served (permissive licence) · get_code("4667b36ac01d0832")
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
import itertools

def get_weights_and_frequencies(N, k=2, l=10):
    f = np.array([[2 * np.pi * i, 2 * np.pi * j] for i, j in itertools.product(range(N), repeat=2)])
    w = np.array([1 / (1 + (2 * np.pi * i / l) ** 2 + (2 * np.pi * j / l) ** 2) ** k for i, j in itertools.product(range(N), repeat=2)]).reshape(-1)
    return (w, f)

ARGS = {
    "N": 3,
    "k": 2,
    "l": 10
}
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
np_random_cos Ran
ran; our contract draft was wrong, not the code
4gnskq5g2s-collab/Astral/datasets/magnetostatics.py
code served (permissive licence) · get_code("1e3b2f680e75fd0a")
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 Add. 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 sympy as sp
import numpy as np

# Analyze the function:
# - x: appears to be indexed as x[0] and x[1], so it needs at least 2 elements
# - amplitudes: has .shape[0] called on it, used in range and indexing, so it's array-like
# - k: numeric parameter with default value 2

# Create small test arguments
x = [sp.Symbol('x_0'), sp.Symbol('x_1')]

# amplitudes should be a numpy array or similar with shape attribute
# N_terms determines the size - let's use 4 terms
amplitudes = np.array([1.0, 2.0, 0.5, 1.5])

k = 2

ARGS = {
    'x': x,
    'amplitudes': amplitudes,
    'k': k
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sympy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Add — a Add output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Add 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
np_random_sin Ran
ran; our contract draft was wrong, not the code
4gnskq5g2s-collab/Astral/datasets/Anisotropic_diffusion.py
code served (permissive licence) · get_code("e1a5662398f9030a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned Add. 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 sympy as sp
import numpy as np
import itertools

# Analyze the function:
# - x: appears to be indexed as x[0] and x[1], so it's a sequence/array-like with at least 2 elements
# - amplitudes: has .shape[0] attribute, so it's a numpy array or similar
# - k: numeric parameter (default 2)
# - eps: numeric parameter (default 1.0)

# The function uses sympy symbols, so x should contain sympy symbols or be compatible
# Looking at the usage: sp.sin(sp.pi * i * x[0]), x[0] and x[1] are multiplied with sympy expressions
# This suggests x should be a list/tuple of sympy symbols

x = [sp.Symbol('x0'), sp.Symbol('x1')]

# amplitudes needs .shape[0] to work, so use numpy array
# N_terms will be determined by amplitudes.shape[0]
# Let's use a small array with 3 terms
amplitudes = np.array([1.0, 0.5, 0.3])

k = 2
eps = 1.0

ARGS = {
    'x': x,
    'amplitudes': amplitudes,
    'k': k,
    'eps': eps
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, sympy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Add — a Add output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Add 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
sigma_11 Not yet run
unverified (harvested, no recorded run)
4gnskq5g2s-collab/Astral/datasets/elastoplasticity.py
code served (permissive licence) · get_code("e58daaccc40792ed")
How it ran

no invocation recorded.

sigma_22 Not yet run
unverified (harvested, no recorded run)
4gnskq5g2s-collab/Astral/datasets/elastoplasticity.py
code served (permissive licence) · get_code("7c404994e2c9b9be")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 12 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
np_random_sin 1 other paper
1 dated earlier
2402.05585 (2024-02)
Neural functional a posteriori error estimates

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 primal approach to physics-informed learning is a residual minimization. We argue that residual is, at best, an indirect measure of the error of approximate solution and propose to train with error majorant instead. Since error majorant provides a direct upper bound on error, one can reliably estimate how close PiNN is to the exact solution and stop the optimization process when the desired accuracy is reached. We call loss function associated with error majorant Astral: neurAl a poSTerioRi functionAl Loss. To compare Astral and residual loss functions, we illustrate how error majorants can be derived for various PDEs and conduct experiments with diffusion equations (including anisotropic and in the L-shaped domain), convection-diffusion equation, temporal discretization of Maxwell's equation, magnetostatics and nonlinear elastoplasticity problems. The results indicate that Astral loss is competitive to the residual loss, typically leading to faster convergence and lower error (e.g., for Maxwell's equations, we observe an order of magnitude better relative error and training time). The main benefit of using Astral loss comes from its ability to estimate error, which is impossible with other loss functions. Our experiments indicate that the error estimate obtained with Astral loss is usually tight enough, e.g., for a highly anisotropic equation, on average, Astral overestimates error by a factor of 1.5, and for convection-diffusion by a factor of 1.7. We further demonstrate that Astral loss is better correlated with error than residual and is a more reliable predictor (in a statistical sense) of the error value. Moreover, unlike residual, the error indicator obtained from Astral loss has a superb spatial correlation with error. Backed with the empirical and theoretical results, we argue that one can productively use Astral loss to perform reliable error analysis and approximate PDE solutions with accuracy similar to standard residual-based techniques.

For agents

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

get_harvested_code_for_paper("2406.02645")
get_code_for_paper("2406.02645")
have("2406.02645")

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

curl https://syntology.ai/api/ran/2406.02645.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.02645.svg)](https://syntology.ai/paper/2406.02645)

Connect an agent — have() is free.