SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2107.00745 · 2021

q-Paths: Generalizing the Geometric Annealing Path using Power Means

arXiv · PDF · Open in the Atlas

Code that ran

13 of 15 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 11 have no contract verdict; 2 failures of our instrument; 2 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/2107.00745.json · badge: /api/ran/2107.00745.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
vmasrani/qpaths_uai_2021 canonical (classifier label)
classifier basis: census
13 ran · 2 no recorded run
FunctionStatusWhere it lives
essl Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/smc/src/qhelpers.py
code served (permissive licence) · get_code("11b3456d762ba64f")
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 float64, ndim 0, 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

# Create a small log-weights array
np.random.seed(0)
lw = np.random.randn(10)

ARGS = {
    "lw": lw
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
gaus_log_prob Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/toy/src/hmc.py
code served (permissive licence) · get_code("e29f668c27639d43")
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 (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

# Construct arguments for gaus_log_prob
x = np.array([1.0, 2.0, 3.0, 4.0])
mean = np.array([0.5, 1.5, 2.5, 3.5])
sig = np.array([0.5, 0.8, 1.2, 1.5])

ARGS = {
    'x': x,
    'mean': mean,
    'sig': sig
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (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
geometric_average Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/toy/src/ais.py
code served (permissive licence) · get_code("13d3e55b78d72af6")
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 Gaussian1D. 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

class Gaussian1D:

    def __init__(self, mean=None, variance=None, precision=None, precisionxmean=None):
        if mean is not None:
            self.mean = mean
            self.variance = variance
            self.precision = 1.0 / variance
            self.precisionxmean = mean / variance
        else:
            self.precision = precision
            self.precisionxmean = precisionxmean
            self.variance = 1.0 / precision
            self.mean = precisionxmean / precision
        self.sigma = np.sqrt(self.variance)

    def sample(self, no_samples):
        return np.random.randn(no_samples) * np.sqrt(self.variance) + self.mean

    def logprob(self, x):
        return -(0.5 * (np.log(2 * np.pi * self.sigma * self.sigma) + ((x - self.mean) / self.sigma) ** 2))

class MVGaussian:

    def __init__(self, mean=None, variance=None, precision=None, precisionxmean=None):
        if mean is not None:
            self.mean = mean
            self.variance = variance
            self.precision = np.linalg.inv(self.variance)
            self.precisionxmean = np.matmul(self.precision, self.mean)
        else:
            self.precision = precision
            self.precisionxmean = precisionxmean
            self.variance = np.linalg.inv(precision)
            self.mean = np.matmul(self.variance, precisionxmean)

    def sample(self, no_samples):
        return np.random.multivariate_normal(self.mean, self.variance, size=(no_samples,))

    def logprob(self, x):
        from scipy.stats import multivariate_normal
        y = multivariate_normal.pdf(x, mean=self.mean, cov=self.variance)
        return y

def geometric_average(dist1, dist2, pow1=0.5, pow2=0.5):
    precision = dist1.precision * pow1 + dist2.precision * pow2
    precisionxmean = dist1.precisionxmean * pow1 + dist2.precisionxmean * pow2
    if isinstance(dist1.precision, float) or isinstance(dist1.precision, int):
        return Gaussian1D(precision=precision, precisionxmean=precisionxmean)
    else:
        return MVGaussian(precision=precision, precisionxmean=precisionxmean)

# Create two 1D Gaussian distributions for testing
dist1 = Gaussian1D(mean=0.0, variance=1.0)
dist2 = Gaussian1D(mean=2.0, variance=1.5)

ARGS = {
    "dist1": dist1,
    "dist2": dist2,
    "pow1": 0.5,
    "pow2": 0.5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Gaussian1D — a Gaussian1D output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Gaussian1D 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
leapfrog Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
vmasrani/qpaths_uai_2021/toy/src/hmc.py
code served (permissive licence) · get_code("7322057dc47258d9")
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

# Define a simple potential function that returns both value and gradient
def potential_fn(q):
    """
    Simple quadratic potential: V(q) = 0.5 * sum(q^2)
    Gradient: dV/dq = q
    """
    V = 0.5 * np.sum(q**2)
    dVdq = np.copy(q)
    return V, dVdq

# Set up small numpy arrays for position and momentum
q_init = np.array([1.0, 2.0, 0.5, -1.5], dtype=np.float64)
p_init = np.array([0.1, -0.2, 0.3, 0.05], dtype=np.float64)

# Compute initial gradient
V_init, dVdq_init = potential_fn(q_init)

# Parameters for the integrator
path_len = 1.0
step_size = 0.1

ARGS = {
    "q": q_init,
    "p": p_init,
    "dVdq": dVdq_init,
    "potential": potential_fn,
    "path_len": path_len,
    "step_size": step_size,
}
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
log_f_i_geometric Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/ais/ais.py
code served (permissive licence) · get_code("00b910b73a0b26b5")
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 (2, 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 torch
import torch.nn as nn

torch.manual_seed(0)

class MockModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 1)
    
    def log_prior(self, z, use_cuda):
        """Return log prior probability of latent z"""
        # Simple mock: sum of squares as log prior
        return -0.5 * (z ** 2).sum(dim=-1, keepdim=True)
    
    def log_likelihood(self, z, data):
        """Return log likelihood of data given latent z"""
        # Simple mock: pass z through linear layer to get likelihood
        batch_size = z.shape[0]
        likelihood = self.linear(z)
        return likelihood

model = MockModel()

# Latent variable: batch_size=2, latent_dim=4
z = torch.randn(2, 4)

# Data: batch_size=2, data_dim=3
data = torch.randn(2, 3)

# Beta parameter: mixture weight between 0 and 1
beta = 0.7

# Use CUDA flag
use_cuda = False

ARGS = {
    "model": model,
    "z": z,
    "data": data,
    "beta": beta,
    "use_cuda": use_cuda
}
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 (2, 1) — shape (2, 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
log_f_i_qpath Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/ais/ais.py
code served (permissive licence) · get_code("4777b5cce48437fa")
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 (2, 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 torch
import torch.nn as nn

torch.manual_seed(0)

class SimpleModel(nn.Module):
    """Minimal model that provides log_prior and log_likelihood methods"""
    
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 1)
    
    def log_prior(self, z, use_cuda):
        """Return log prior as a tensor"""
        # Simple Gaussian prior: -0.5 * sum(z^2)
        return -0.5 * torch.sum(z ** 2, dim=-1, keepdim=True)
    
    def log_likelihood(self, z, data):
        """Return log likelihood as a tensor"""
        # Simple likelihood based on z and data
        # data shape: (2, 3), z shape: (2, 4)
        x, y = data
        # Project z to match data dimension
        pred = self.linear(z)  # (2, 1)
        # Simple Gaussian likelihood
        return -0.5 * torch.sum((y - pred) ** 2, dim=-1, keepdim=True)

model = SimpleModel()

# Create latent variable z: shape (2, 4)
z = torch.randn(2, 4)

# Create data as tuple (x, y)
x = torch.randn(2, 3)
y = torch.randn(2, 1)
data = (x, y)

# Other parameters
beta = 0.5
q = 0.8
use_cuda = False

ARGS = {
    "model": model,
    "z": z,
    "data": data,
    "beta": beta,
    "q": q,
    "use_cuda": use_cuda
}
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 (2, 1) — shape (2, 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
log_sum_weighted_exp Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/ais/ais.py
code served (permissive licence) · get_code("2fd96fd053ae2bbd")
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 (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create small tensors for the function parameters
# val1 and val2 should be comparable tensors (same shape)
val1 = torch.randn(2, 4)
val2 = torch.randn(2, 4)

# weight1 and weight2 should be scalars or broadcastable tensors
weight1 = torch.tensor(0.3)
weight2 = torch.tensor(0.7)

ARGS = {
    "val1": val1,
    "val2": val2,
    "weight1": weight1,
    "weight2": weight2,
}
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 (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
log_sum_weighted_exp Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/smc/src/qhelpers.py
code served (permissive licence) · get_code("d53b94e96c4b67fd")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Construct arguments for log_sum_weighted_exp
# The function takes 4 numpy array parameters: val1, val2, weight1, weight2
# It performs element-wise operations, so all should have compatible shapes

val1 = np.array([[1.0, 2.0, 3.0, 4.0],
                 [0.5, 1.5, 2.5, 3.5]])

val2 = np.array([[0.8, 1.9, 2.8, 3.9],
                 [0.3, 1.2, 2.1, 3.0]])

weight1 = np.array([[0.6, 0.6, 0.6, 0.6],
                    [0.6, 0.6, 0.6, 0.6]])

weight2 = np.array([[0.4, 0.4, 0.4, 0.4],
                    [0.4, 0.4, 0.4, 0.4]])

ARGS = {
    'val1': val1,
    'val2': val2,
    'weight1': weight1,
    'weight2': weight2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
logpdf_to_qlogpdf Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/smc/src/qhelpers.py
code served (permissive licence) · get_code("9176ce140edef3ad")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# logpdf_to_qlogpdf takes two parameters:
# 1. logpdf: a numpy array of log probability density values
# 2. q: a scalar parameter (likely a float)

# The function computes: (np.exp((1 - q) * logpdf) - 1) / (1 - q)
# This requires:
# - logpdf to be a numpy array that can be multiplied by a scalar
# - q to be a scalar value (float or int)

# Create a small logpdf array (batch size 2, features 4)
logpdf_value = np.random.randn(2, 4) - 5  # negative values typical for log PDFs

# Create a q parameter (scalar, must not equal 1 to avoid division by zero)
q_value = 0.5

ARGS = {
    "logpdf": logpdf_value,
    "q": q_value
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
pmap Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/toy/src/utils.py
code served (permissive licence) · get_code("d4624bf8815e34ed")
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
from tqdm import tqdm

torch.manual_seed(0)

def simple_function(x):
    """A simple function to parallelize"""
    return x * 2

def another_function(item):
    """Another simple function that processes items"""
    return item + 10

# Define the ARGS dictionary with all parameters for pmap
ARGS = {
    'f': simple_function,
    'arr': [1, 2, 3, 4, 5],
    'n_jobs': 2,
    'notebook': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
joblib, numpy, torch, tqdm — 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
student_moment_average Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/toy/src/ais.py
code served (permissive licence) · get_code("edb6b4ae38f4ff33")
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 Student1D. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from scipy import stats
from autograd.scipy import stats as autograd_stats

class Student1D:

    def __init__(self, q=None, mean=None, variance=None, precision=None, precisionxmean=None, df=None):
        self.df = (3 - q) / (q - 1) if q is not None and q != 1 else df
        if self.df is None:
            raise ValueError('Please specify q parameter (q!=1) or df (degrees of freedom)')
        self.q = q if q is not None else (self.df + 3) / (self.df + 1)
        if mean is not None:
            self.mean = mean
            self.variance = variance
            self.precision = 1.0 / variance
            self.precisionxmean = mean / variance
        elif precisionxmean is not None:
            self.precision = precision
            self.precisionxmean = precisionxmean
            self.variance = 1.0 / precision
            self.mean = precisionxmean / precision
        else:
            self.mean = 0
            self.variance = 1
            self.precision = 1.0 / variance
            self.precisionxmean = mean / variance
        self.sigma = np.sqrt(self.variance)
        self.student_t = stats.t
        self.student_t_autograd = autograd_stats.t

    def sample(self, no_samples):
        return self.student_t.rvs(self.df, size=no_samples, loc=self.mean, scale=self.sigma)

    def logprob(self, x):
        return self.student_t_autograd.logpdf(x, self.df, loc=self.mean, scale=self.sigma)

def student_moment_average(dist1, dist2, beta):
    pow1 = 1 - beta
    pow2 = beta
    t_scaling = (dist2.df + 2) / dist2.df
    t_mean = pow1 * dist1.mean + pow2 * dist2.mean
    t_var = pow1 * dist1.variance + pow2 * dist2.variance + t_scaling * pow1 * pow2 * np.outer(dist2.mean - dist1.mean, dist2.mean - dist1.mean)
    return Student1D(df=dist2.df, mean=t_mean, variance=t_var)

# Create two Student1D distributions
dist1 = Student1D(df=5.0, mean=np.array(1.0), variance=np.array(2.0))
dist2 = Student1D(df=8.0, mean=np.array(0.5), variance=np.array(1.5))

# Create beta parameter for mixing
beta = 0.3

ARGS = {
    'dist1': dist1,
    'dist2': dist2,
    'beta': beta
}
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 Student1D — a Student1D output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Student1D 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
tqdm_joblib Ran
ran; our contract draft was wrong, not the code
vmasrani/qpaths_uai_2021/toy/src/utils.py
code served (permissive licence) · get_code("fc7032ef481eaeb0")
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 _GeneratorContextManager. 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 contextlib
import joblib
from tqdm import tqdm

# Create a simple tqdm object that the context manager will use
tqdm_obj = tqdm(total=100, desc="Processing")

ARGS = {
    "tqdm_object": tqdm_obj
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
joblib, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type _GeneratorContextManager — a _GeneratorContextManager output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a _GeneratorContextManager 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
validate_dataset_path Ran
ran on a synthesized input
vmasrani/qpaths_uai_2021/smc/src/assertions.py
code served (permissive licence) · get_code("a076b5ac757f4ad0")
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

# Construct a minimal args object
class Args:
    def __init__(self):
        self.data_dir = '/tmp/test_data/'

ARGS = {'args': Args()}
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
hmc Not yet run
unverified (harvested, no recorded run)
vmasrani/qpaths_uai_2021/toy/src/hmc.py
code served (permissive licence) · get_code("6ab30bbf0cd034ba")
How it ran

no invocation recorded.

moment_average Not yet run
unverified (harvested, no recorded run)
vmasrani/qpaths_uai_2021/toy/src/ais.py
code served (permissive licence) · get_code("554153c2f7d049eb")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 2 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 1 later) and 13 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
leapfrog 1 other paper
1 dated earlier
1810.04449 (2018-10)
Faster Hamiltonian Monte Carlo by Learning Leapfrog Scale: a self-calibrated randomized solution
tqdm_joblib 1 other paper
1 later
2110.10713 (2021-10)
PPFS: Predictive Permutation Feature Selection

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

Many common machine learning methods involve the geometric annealing path, a sequence of intermediate densities between two distributions of interest constructed using the geometric average. While alternatives such as the moment-averaging path have demonstrated performance gains in some settings, their practical applicability remains limited by exponential family endpoint assumptions and a lack of closed form energy function. In this work, we introduce $q$-paths, a family of paths which is derived from a generalized notion of the mean, includes the geometric and arithmetic mixtures as special cases, and admits a simple closed form involving the deformed logarithm function from nonextensive thermodynamics. Following previous analysis of the geometric path, we interpret our $q$-paths as corresponding to a $q$-exponential family of distributions, and provide a variational representation of intermediate densities as minimizing a mixture of $α$-divergences to the endpoints. We show that small deviations away from the geometric path yield empirical gains for Bayesian inference using Sequential Monte Carlo and generative model evaluation using Annealed Importance Sampling.

For agents

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

get_harvested_code_for_paper("2107.00745")
get_code_for_paper("2107.00745")
have("2107.00745")

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

curl https://syntology.ai/api/ran/2107.00745.json

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

[![Syntology run record](https://syntology.ai/api/ran/2107.00745.svg)](https://syntology.ai/paper/2107.00745)

Connect an agent — have() is free.