SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2004.04096 · 2020

Probabilistic embeddings for speaker diarization

arXiv · PDF · Open in the Atlas

Code that ran

19 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 19 have no contract verdict; 0 failures of our instrument; 2 no recorded run), measured 2026-11-21; 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/2004.04096.json · badge: /api/ran/2004.04096.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
bsxfan/probabilistic_embeddings canonical (classifier label)
classifier basis: census
19 ran · 2 no recorded run
FunctionStatusWhere it lives
PartitionsAsBlocklabels Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/combin/partitions.py
code served (permissive licence) · get_code("84e5da8d52aeb6f6")
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 map. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

ARGS = {
    'n': 4
}
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 map — a map output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a map 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
blocklabels_to_flagvector Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/combin/partitions.py
code served (permissive licence) · get_code("ff70ee27c442a857")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 1, dtype family int, shape (8,) 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

# Understand the function:
# - labels: m-vector with entries in {0,...,2**n-1}
#   These are indices into the output array
# - sz: must be 2**n, the size of the output array
# - dtype: data type for the output
#
# The function creates a zero array of size sz with given dtype,
# then sets r[labels] = 1, so labels must be valid indices into an array of size sz.
#
# Example: if sz=8 (so n=3), labels could be [0, 2, 5] (3 blocks)
# Result would be a vector of length 8 with 1s at positions 0, 2, 5

# Create valid arguments
labels = np.array([0, 2, 5], dtype=np.int64)  # 3 blocks
sz = 8  # 2**3 = 8 elements in the partitioned set
dtype = int

ARGS = {
    "labels": labels,
    "sz": sz,
    "dtype": dtype
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (8,) — shape (8,) 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
cs Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/deriv/adtools.py
code served (permissive licence) · get_code("2f2a728ee2342d83")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

def simple_func(x):
    """A simple function that works with complex step derivative."""
    return np.sin(x) + 2 * x**2

def array_func(x):
    """A function that works elementwise on arrays."""
    return np.exp(x) - np.cos(x)

# Test with scalar
scalar_x = 1.5

# Test with array
array_x = np.array([0.1, 0.2, 0.3, 0.5])

ARGS = {
    "f": simple_func,
    "x": scalar_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 tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
cs_erfinv Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/special/erf_tools.py
code served (permissive licence) · get_code("1968757262788202")
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
from scipy import special

# Test with both real and complex inputs
# The function accepts a single parameter p which can be real or complex

# Real scalar
real_scalar = 0.5

# Complex scalar
complex_scalar = 0.3 + 0.1j

# Real array
real_array = np.array([0.1, 0.5, 0.9])

# Complex array
complex_array = np.array([0.2 + 0.05j, 0.5 + 0.1j, 0.8 + 0.02j])

# Using a real scalar as the main test case
ARGS = {
    "p": real_scalar
}
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 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
erf Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/special/erf_tools.py
code served (permissive licence) · get_code("d504382c533f0747")
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
from scipy import special

sqrtpi = np.sqrt(np.pi)

def erf(x, deriv=False):
    y = special.erf(x)
    if not deriv:
        return y
    return (y, lambda dy: 2.0 / sqrtpi * dy * np.exp(-x ** 2))

# Setup: construct valid arguments
x = np.array([0.5, 1.0, -0.3, 0.0])
deriv = False

ARGS = {
    "x": x,
    "deriv": deriv
}
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 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
erf_and_logderiv Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/special/erf_tools.py
code served (permissive licence) · get_code("3305ad9c04c0c4ba")
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
from scipy import special

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

# Create test inputs
x = np.array([0.5, 1.2, -0.3, 2.1], dtype=np.float64)

# Define ARGS dictionary with all parameters
ARGS = {
    'x': x,
    'deriv': False,
}
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 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
k012e Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/special/besselk.py
code served (permissive licence) · get_code("43809375625a7a32")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import scipy.special as special
import numpy as np

# Test with a real scalar value
real_x = 2.5

# Test with a complex scalar value
complex_x = 2.5 + 1.0j

# Test with a real array
real_array = np.array([1.0, 2.0, 3.0, 4.0])

# Test with a complex array
complex_array = np.array([1.0 + 0.5j, 2.0 + 1.0j, 3.0 + 1.5j, 4.0 + 2.0j])

# Create ARGS dict with different test cases
ARGS = {
    "x": real_x
}
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 tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
k0e Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/special/besselk.py
code served (permissive licence) · get_code("df374dee154cb865")
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 scipy.special as special
import numpy as np

# The function k0e(x) takes a single parameter x
# It checks if x is complex using np.iscomplexobj(x)
# If complex, it calls special.kve(0, x)
# If not complex, it calls special.k0e(x)
# Both scipy functions accept array-like inputs

# Test with a real-valued numpy array
x_real = np.array([0.5, 1.0, 1.5, 2.0])

ARGS = {
    "x": x_real
}
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 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
k1e Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/special/besselk.py
code served (permissive licence) · get_code("9ef74bb813161dd7")
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 scipy.special as special
import numpy as np
import torch

torch.manual_seed(0)

# The function k1e takes a single parameter x
# It checks if x is a complex object, then calls either special.kve(1, x) or special.k1e(x)
# Both scipy special functions accept array-like inputs

# Test with real numpy array
x_real = np.array([0.5, 1.0, 2.0, 3.0])

ARGS = {
    "x": x_real
}
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 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
labels_to_1hotmatrix Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/combin/partitions.py
code served (permissive licence) · get_code("5d6e193b04f14b0a")
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, 6) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create a valid restricted growth string
# A restricted growth string has:
# - first entry is 0
# - each subsequent entry cannot exceed the max of previous entries by more than 1
labels = np.array([0, 1, 1, 2, 2, 1], dtype=int)

ARGS = {
    "labels": labels,
    "dtype": int
}
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 int, shape (3, 6) — shape (3, 6) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
logprob_alpha Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/prob/crp.py
code served (permissive licence) · get_code("b89ae8ae46f97e4e")
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
from scipy.special import gammaln, psi

np.random.seed(0)

# Analyze the function:
# - alpha: a positive scalar (float)
# - counts: a sequence/array where len(counts) gives K (number of categories)
#           and sum(counts) gives T (total count)
#           Individual elements are used with gammaln(counts) - needs to be numeric

# Create small valid arguments
alpha = 2.5  # positive scalar

# counts should be a sequence of counts
# Let's make K=4 distinct categories with small counts
counts = [1, 2, 3, 2]  # K=4, T=8

ARGS = {
    "alpha": alpha,
    "counts": counts,
}
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 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
logprob_alpha_beta Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/prob/crp.py
code served (permissive licence) · get_code("1d37d674cf3f39f3")
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

# Construct valid arguments for logprob_alpha_beta
# The function expects:
# - alpha: float > 0
# - beta: float where 0 < beta < 1
# - counts: list/array of counts (non-negative integers)

# Create valid parameters
alpha = 2.5  # Must be > 0
beta = 0.3   # Must be 0 < beta < 1
counts = np.array([2, 3, 1, 4])  # Array of counts, all positive integers

# Verify constraints before building ARGS
assert alpha > 0, "alpha must be > 0"
assert 0 < beta < 1, "beta must be between 0 and 1 (exclusive)"
assert all(c > beta for c in counts), "all counts must be > beta for gammaln(counts - beta) to be valid"

ARGS = {
    "alpha": alpha,
    "beta": beta,
    "counts": counts
}
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 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
logsumexp Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/deriv/adfunctions.py
code served (permissive licence) · get_code("f64ca25c9836a19a")
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 (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

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

# Create a small input array for logsumexp
# Shape: (2, 3) - 2 samples, 3 dimensions
X = np.random.randn(2, 3).astype(np.float64)

# Define ARGS with all parameters
ARGS = {
    "X": X,
    "axis": 0,
    "keepdims": False,
    "deriv": False,
}
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 (3,) — shape (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
raug Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/deriv/adtools.py
code served (permissive licence) · get_code("8f2cd7ed3c038987")
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 complex, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Test case 1: scalar inputs
x_scalar = 5.0
r_scalar = 2.0

# Test case 2: numpy array inputs
x_array = np.array([[1.0, 2.0, 3.0, 4.0],
                     [5.0, 6.0, 7.0, 8.0]])
r_array = np.array([[0.1, 0.2, 0.3, 0.4],
                     [0.5, 0.6, 0.7, 0.8]])

# Test case 3: tuple of scalars
x_tuple_scalar = (1.0, 2.0, 3.0)
r_tuple_scalar = (0.1, 0.2, 0.3)

# Test case 4: tuple of arrays
x_array1 = np.array([1.0, 2.0, 3.0, 4.0])
r_array1 = np.array([0.1, 0.2, 0.3, 0.4])
x_array2 = np.array([5.0, 6.0, 7.0, 8.0])
r_array2 = np.array([0.5, 0.6, 0.7, 0.8])
x_tuple_array = (x_array1, x_array2)
r_tuple_array = (r_array1, r_array2)

# Test case 5: nested tuple structure
x_nested = (1.0, (2.0, 3.0))
r_nested = (0.1, (0.2, 0.3))

# Using scalar case as primary test
ARGS = {
    'x': x_scalar,
    'r': r_scalar,
    'eps': 1e-20j
}
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 complex, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rdot Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/deriv/adtools.py
code served (permissive licence) · get_code("4d6df300e2ac88a3")
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 float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Test case 1: two scalars
x_scalar = 3.5
y_scalar = 2.0

# Test case 2: two ndarrays with same shape
x_array = np.array([1.0, 2.0, 3.0, 4.0])
y_array = np.array([5.0, 6.0, 7.0, 8.0])

# Test case 3: nested tuple structure with scalars and arrays
x_nested = (
    1.5,
    np.array([1.0, 2.0]),
    (
        2.5,
        np.array([[1.0, 2.0], [3.0, 4.0]])
    )
)

y_nested = (
    2.5,
    np.array([3.0, 4.0]),
    (
        3.5,
        np.array([[5.0, 6.0], [7.0, 8.0]])
    )
)

# Use the scalar case as the main test case
ARGS = {
    'x': x_scalar,
    'y': y_scalar
}
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 float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
softplus Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/special/softplus.py
code served (permissive licence) · get_code("d82cc22f6cca675b")
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

def softplus(x):
    """log( 1 + exp(x) )"""
    return np.log1p(np.exp(-np.abs(x))) + np.maximum(x, 0)

# Create test input - a small numpy array
x = np.array([[-2.5, -1.0, 0.0, 1.5],
              [2.0, -0.5, 0.5, 3.0]], dtype=np.float32)

ARGS = {
    '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, 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
softplusinv Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/special/softplus.py
code served (permissive licence) · get_code("e4239541a6ea39c3")
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 a small numpy array to pass as y
# The function expects a numpy array (or array-like) and performs mathematical operations
y = np.array([0.5, 1.0, 2.0, -0.5])

ARGS = {
    "y": y
}
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
squeeze Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/deriv/adfunctions.py
code served (permissive licence) · get_code("e56c64810abf04fe")
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 (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

# Test case 1: 2D array with axis=0
A1 = np.array([[1, 2, 3], [4, 5, 6]])

# Test case 2: 3D array with axis=1
A2 = np.array([[[1, 2], [3, 4]], [[5, 6], [7, 8]]])

# Test case 3: array with singleton dimension to squeeze to scalar
A3 = np.array([[[42]]])

# Use the first test case for the main ARGS
ARGS = {
    "A": A1,
    "axis": None
}
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 int, 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
sum2shape Ran
ran on a synthesized input
bsxfan/probabilistic_embeddings/lib/deriv/broadcast_adtools.py
code served (permissive licence) · get_code("f606a62d87466fd7")
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, 2) 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

# Test case 1: x already matches to_shape
x1 = np.array([[1, 2], [3, 4]])
to_shape1 = (2, 2)

# Test case 2: x needs to be summed to match to_shape (scalar output)
x2 = np.array([[1, 2, 3], [4, 5, 6]])
to_shape2 = ()

# Test case 3: x needs to be summed along one axis to match to_shape
x3 = np.random.randn(4, 1, 3)
to_shape3 = (4, 1, 3)

# Test case 4: x needs broadcasting-like summing
x4 = np.random.randn(2, 3, 4)
to_shape4 = (2, 1, 4)

# Use test case 1 as the primary example
ARGS = {
    "x": np.array([[1.0, 2.0], [3.0, 4.0]]),
    "to_shape": (2, 2)
}
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, 2) — shape (2, 2) 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
logprob_beta Not yet run
unverified (harvested, no recorded run)
bsxfan/probabilistic_embeddings/lib/prob/crp.py
code served (permissive licence) · get_code("a6571a54ed4fb4d9")
How it ran

no invocation recorded.

sum2sh Not yet run
unverified (harvested, no recorded run)
bsxfan/probabilistic_embeddings/lib/deriv/broadcast_adtools.py
code served (permissive licence) · get_code("3bcb697e90534a24")
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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Speaker embeddings (x-vectors) extracted from very short segments of speech have recently been shown to give competitive performance in speaker diarization. We generalize this recipe by extracting from each speech segment, in parallel with the x-vector, also a diagonal precision matrix, thus providing a path for the propagation of information about the quality of the speech segment into a PLDA scoring backend. These precisions quantify the uncertainty about what the values of the embeddings might have been if they had been extracted from high quality speech segments. The proposed probabilistic embeddings (x-vectors with precisions) are interfaced with the PLDA model by treating the x-vectors as hidden variables and marginalizing them out. We apply the proposed probabilistic embeddings as input to an agglomerative hierarchical clustering (AHC) algorithm to do diarization in the DIHARD'19 evaluation set. We compute the full PLDA likelihood 'by the book' for each clustering hypothesis that is considered by AHC. We do joint discriminative training of the PLDA parameters and of the probabilistic x-vector extractor. We demonstrate accuracy gains relative to a baseline AHC algorithm, applied to traditional xvectors (without uncertainty), and which uses averaging of binary log-likelihood-ratios, rather than by-the-book scoring.

For agents

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

get_harvested_code_for_paper("2004.04096")
get_code_for_paper("2004.04096")
have("2004.04096")

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

curl https://syntology.ai/api/ran/2004.04096.json

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

[![Syntology run record](https://syntology.ai/api/ran/2004.04096.svg)](https://syntology.ai/paper/2004.04096)

Connect an agent — have() is free.