SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2302.12247 · NeurIPS · 2023

Quantifying & Modeling Multimodal Interactions: An Information Decomposition Framework

Paul Liang, Nicholas Allen, Randy Auerbach, Ruslan Salakhutdinov, Louis-Philippe Morency, Cmu, Richard Chen, Xiang Fan, Yun Cheng, Zihao Deng, Chun Ling, Suzanne Nie, and 1 more

arXiv · PDF · Open in the Atlas

Code that ran

15 of 16 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, 1 violated it and 0 have no contract verdict; 14 failures of our instrument; 1 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/2302.12247.json · badge: /api/ran/2302.12247.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
pliang279/pid canonical (classifier label)
classifier basis: classifier_2025
15 ran · 1 no recorded run
FunctionStatusWhere it lives
CI Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("da154993a013ef1d")
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 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 two 3D numpy arrays with distinct dimensions
# Shape: (height=2, width=3, channels=4)
P = np.random.rand(2, 3, 4)
Q = np.random.rand(2, 3, 4)

# Normalize to make them valid probability distributions
P = P / P.sum()
Q = Q / Q.sum()

ARGS = {
    "P": P,
    "Q": Q,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
CoI Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("ba0c1a2505d20c59")
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 valid 3D probability distribution P with shape (X1, X2, Y)
# Using distinct sizes: 4, 3, 2 for each dimension
np.random.seed(0)
P = np.random.dirichlet(np.ones(4 * 3 * 2)).reshape((4, 3, 2))

# Normalize to ensure it's a valid probability distribution
P = P / P.sum()

ARGS = {
    "P": P
}
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
MI Ran
ran, violated the contract we drafted
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("5dd04c930a0883b6")
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 2D probability matrix
# Using shape (3, 4) to have distinct dimensions
P = np.array([
    [0.1, 0.15, 0.2, 0.05],
    [0.05, 0.1, 0.15, 0.1],
    [0.1, 0.05, 0.05, 0.1]
], dtype=np.float64)

# Normalize to ensure it's a valid probability distribution
P = P / P.sum()

ARGS = {
    "P": P
}
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
UI Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("3a6eae326a67768e")
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 valid 3D probability distribution
# Shape: (X1_dim=4, X2_dim=3, Y_dim=2)
# The function uses:
# - P.sum(axis=(1, 2)) for cond_id=0, resulting in shape (4,)
# - P.sum(axis=(0, 2)) for cond_id=1, resulting in shape (3,)
# - P[i, :, :] slicing for cond_id=0
# - P[:, i, :] slicing for cond_id=1
# - Normalization by dividing slices by their sum
# - MI() function which expects 2D arrays

# Create a random 3D array and normalize to form a valid probability distribution
np.random.seed(0)
P_raw = np.random.rand(4, 3, 2)
P = P_raw / P_raw.sum()  # Normalize so all entries sum to 1

ARGS = {
    "P": P,
    "cond_id": 0,
}
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
convert_data_to_distribution Ran
ran; our contract draft was wrong, not the code
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("32886135fe06536a")
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 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

# Create small numpy arrays with categorical data
# x1: 6 samples with values from {0, 1, 2}
x1 = np.array([0, 1, 2, 0, 1, 2])

# x2: 6 samples with values from {10, 20}
x2 = np.array([10, 20, 10, 20, 10, 20])

# y: 6 samples with values from {100, 200, 300}
y = np.array([100, 200, 300, 100, 200, 300])

ARGS = {
    "x1": x1,
    "x2": x2,
    "y": y,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
extract_categorical_from_data Ran
ran; our contract draft was wrong, not the code
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("d30928c06a648d94")
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 tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# - extract_categorical_from_data takes parameter 'x'
# - It calls set(x), iterates over x, and indexes into x with x_
# - This means x should be an iterable of hashable items
# - The function maps raw values to discrete integer indices

# Create a simple list with categorical/hashable values
x = ['cat', 'dog', 'cat', 'bird', 'dog', 'cat']

ARGS = {
    'x': x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — 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_bounds Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("6de22f9453a75297")
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 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
from scipy.special import rel_entr
from scipy.stats import entropy

# Create a small valid 3D probability distribution P
# Dimensions: X1 (size 3), X2 (size 4), Y (size 2)
np.random.seed(0)
P_raw = np.random.rand(3, 4, 2)
P = P_raw / P_raw.sum()  # Normalize to be a valid probability distribution

ARGS = {
    "P": P
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
cvxpy, 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
get_quantities Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("3a7e38eff2e450b8")
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 numpy as np
from scipy.special import rel_entr
from scipy.stats import entropy

# Create a small valid 3D probability distribution P
# Dimensions: x1=2, x2=3, y=4
np.random.seed(0)
P_raw = np.random.rand(2, 3, 4)
# Normalize to make it a valid probability distribution (sum to 1)
P = P_raw / P_raw.sum()

ARGS = {
    "P": P
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cvxpy, numpy, scipy — 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
majorization_bound Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("31bbe15992d2ab17")
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.stats import entropy

# Construct arguments for majorization_bound function
# The function takes two parameters: Px1x2_ and Py_
# Both should be array-like objects (numpy arrays) that can be flattened, sorted, and padded

# Create a small probability-like distribution for Px1x2_
# This represents a joint probability distribution that will be flattened
Px1x2_value = np.array([0.3, 0.2, 0.25, 0.15, 0.1])

# Create a small probability-like distribution for Py_
# This represents a marginal probability distribution
Py_value = np.array([0.4, 0.35, 0.25])

ARGS = {
    "Px1x2_": Px1x2_value,
    "Py_": Py_value
}
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
simple_discrim Ran
ran; our contract draft was wrong, not the code
pliang279/pid/estimators/ce_alignment_information.py
code served (permissive licence) · get_code("235a0f630ddd0692")
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 function. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Analyze the function:
# - xs: list of tensors, each has .size(1) called on it (gets dimension 1)
#   The function does torch.argmax(x[i]) for each x in xs, so xs is a list of 2D tensors
#   shape[i] will be the size of dimension 1 for each tensor
# - y: tensor where y[i].item() is called, so it's a 1D tensor of integers
# - num_labels: integer for the number of labels

# Create xs: list of 2D tensors with distinct sizes for dimension 1
# Batch size is 2, dimension 1 sizes are 4 and 3 (distinct)
x1 = torch.randn(2, 4)  # batch_size=2, feature_dim=4
x2 = torch.randn(2, 3)  # batch_size=2, feature_dim=3
xs = [x1, x2]

# Create y: 1D tensor of label indices (batch_size=2)
# Labels should be in range [0, num_labels)
num_labels = 5
y = torch.tensor([0, 2], dtype=torch.long)

ARGS = {
    "xs": xs,
    "y": y,
    "num_labels": num_labels
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
matplotlib, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
sinkhorn_probs Ran
ran; our contract draft was wrong, not the code
pliang279/pid/estimators/ce_alignment_information.py
code served (permissive licence) · get_code("e84664190f35139b")
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 tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create small test tensors
# matrix: 2D tensor that will be normalized
matrix = torch.randn(3, 4)

# x1_probs: probability distribution for dimension 0 (size 3)
x1_probs = torch.tensor([0.2, 0.3, 0.5])

# x2_probs: probability distribution for dimension 1 (size 4)
x2_probs = torch.tensor([0.1, 0.2, 0.3, 0.4])

ARGS = {
    "matrix": matrix,
    "x1_probs": x1_probs,
    "x2_probs": x2_probs,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
matplotlib, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
solve_Q Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("99576a1777b1c03f")
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 (4, 3, 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

# Create a small 3D probability array P with dimensions (x1, x2, y)
# Using distinct sizes: 4, 3, 2 for x1, x2, y respectively
np.random.seed(0)
P = np.random.dirichlet(np.ones(4 * 3 * 2)).reshape((4, 3, 2))

ARGS = {
    "P": P
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cvxpy, 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 (4, 3, 2) — shape (4, 3, 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
solve_Q_other Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("5a0f6afd9da79b06")
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

# Create a small 3D probability array P with dimensions (x1, x2, y)
# Using distinct sizes: 4 for x1, 3 for x2, 2 for y
np.random.seed(0)
P = np.random.dirichlet(np.ones(4 * 3 * 2)).reshape(4, 3, 2)

# Ensure P is a valid probability distribution
P = P / P.sum()

ARGS = {
    "P": P
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cvxpy, 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
test Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("1eed151775ba41d8")
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

# Create a valid 3D probability distribution P
# P has shape (x1_dim, x2_dim, y_dim) = (3, 4, 2)
# We need P to be a valid joint probability distribution that sums to 1

np.random.seed(0)
P_raw = np.random.rand(3, 4, 2)
P = P_raw / P_raw.sum()  # Normalize to sum to 1

ARGS = {
    "P": P
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cvxpy, 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
upper_bound Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
pliang279/pid/bounds/bounds.py
code served (permissive licence) · get_code("248d63471d1f9b6d")
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.stats import entropy

# First, let's understand what upper_bound needs:
# - P: a 3D array that gets sliced with np.sum operations
# - all_p: a dict with keys 'H_x1x2', 'H_y', 'R', 'U1', 'U2'

# Create a small 3D probability distribution array
# P should have shape (x1_size, x2_size, y_size)
# Using distinct dimensions: 4, 3, 2
np.random.seed(0)
P = np.random.dirichlet(np.ones(2), size=(4, 3)).reshape(4, 3, 2)
P = P / P.sum()  # Normalize to be a valid probability distribution

# Create the all_p dictionary with required keys
all_p = {
    'H_x1x2': 2.5,  # entropy value
    'H_y': 1.8,     # entropy value
    'R': 0.5,       # rate
    'U1': 0.3,      # utility 1
    'U2': 0.2       # utility 2
}

ARGS = {
    'P': P,
    'all_p': all_p
}
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
mlp Not yet run
unverified (harvested, no recorded run)
pliang279/pid/estimators/ce_alignment_information.py
code served (permissive licence) · get_code("a1dadeb1ac1d82ee")
How it ran

no invocation recorded. this sample carries JSON test cases instead of a setup block; get_reference_implementation serves them with the level they back.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 5 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 5 later) and 11 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
mlp 3 other papers
1 dated earlier, 2 later
2006.05553 (2020-06)
Neural Methods for Point-wise Dependency Estimation
2306.04539 (2023-06)
Multimodal Learning Without Labeled Multimodal Data: Guarantees and Applications
2603.29676 (2026-03)
A Comprehensive Information-Decomposition Analysis of Large Vision-Language Models
convert_data_to_distribution 2 other papers
2 later
2306.04125 (2023-06)
Multimodal Fusion Interactions: A Study of Human and Automatic Quantification
2306.04539 (2023-06)
Multimodal Learning Without Labeled Multimodal Data: Guarantees and Applications
extract_categorical_from_data 2 other papers
2 later
2306.04125 (2023-06)
Multimodal Fusion Interactions: A Study of Human and Automatic Quantification
2306.04539 (2023-06)
Multimodal Learning Without Labeled Multimodal Data: Guarantees and Applications
sinkhorn_probs 2 other papers
2 later
2306.04539 (2023-06)
Multimodal Learning Without Labeled Multimodal Data: Guarantees and Applications
2603.29676 (2026-03)
A Comprehensive Information-Decomposition Analysis of Large Vision-Language Models
simple_discrim 1 other paper
1 later
2306.04539 (2023-06)
Multimodal Learning Without Labeled Multimodal Data: Guarantees and Applications

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 recent explosion of interest in multimodal applications has resulted in a wide selection of datasets and methods for representing and integrating information from different modalities. Despite these empirical advances, there remain fundamental research questions: How can we quantify the interactions that are necessary to solve a multimodal task? Subsequently, what are the most suitable multimodal models to capture these interactions? To answer these questions, we propose an information-theoretic approach to quantify the degree of redundancy, uniqueness, and synergy relating input modalities with an output task. We term these three measures as the PID statistics of a multimodal distribution (or PID for short), and introduce two new estimators for these PID statistics that scale to high-dimensional distributions. To validate PID estimation, we conduct extensive experiments on both synthetic datasets where the PID is known and on large-scale multimodal benchmarks where PID estimations are compared with human annotations. Finally, we demonstrate their usefulness in (1) quantifying interactions within multimodal datasets, (2) quantifying interactions captured by multimodal models, (3) principled approaches for model selection, and (4) three real-world case studies engaging with domain experts in pathology, mood prediction, and robotic perception where our framework helps to recommend strong multimodal models for each application.

For agents

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

get_harvested_code_for_paper("2302.12247")
get_code_for_paper("2302.12247")
have("2302.12247")

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

curl https://syntology.ai/api/ran/2302.12247.json

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

[![Syntology run record](https://syntology.ai/api/ran/2302.12247.svg)](https://syntology.ai/paper/2302.12247)

Connect an agent — have() is free.