SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2105.06138 · IJCAI · 2021

Unsupervised Hashing with Contrastive Information Bottleneck

Changyou Chen, Zijing Ou, Jianxing Yu, Qinliang Su, Zexuan Qiu

arXiv · PDF · Open in the Atlas

Code that ran

11 of 15 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (6 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 1 violated it and 7 have no contract verdict; 3 failures of our instrument; 4 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2105.06138.json · badge: /api/ran/2105.06138.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
qiuzx2/CIBHash community
Papers with Code archive
11 ran · 4 no recorded run
FunctionStatusWhere it lives
GaussianBlur Ran
ran on a synthesized input · constructed an object; no result was computed
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("545219e390d7b28b")
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 an instance of its own class, GaussianBlur: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import cv2

class GaussianBlur(object):

    def __init__(self, kernel_size, min=0.1, max=2.0):
        self.min = min
        self.max = max
        self.kernel_size = kernel_size

    def __call__(self, sample):
        sample = np.array(sample)
        prob = np.random.random_sample()
        if prob < 0.5:
            sigma = (self.max - self.min) * np.random.random_sample() + self.min
            sample = cv2.GaussianBlur(sample, (self.kernel_size, self.kernel_size), sigma)
        return sample


# Create a small sample image (4x3 RGB)
sample_image = np.random.randint(0, 256, (4, 3, 3), dtype=np.uint8)

ARGS = {
    'kernel_size': 5,
    'min': 0.1,
    'max': 2.0,
}

# Test instantiation
blur_transform = GaussianBlur(**ARGS)
result = blur_transform(sample_image)
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type GaussianBlur — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a GaussianBlur 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
Logger Ran
ran on a synthesized input · constructed an object; no result was computed
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("a5f55809f0ced9ca")
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 an instance of its own class, Logger: the run constructed the object and did not call it, so no result was computed. 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 math
import os
import statistics as stat
import sys
import tempfile

ARGS = {
    'log_path': os.path.join(tempfile.gettempdir(), 'test_log.txt'),
    'on': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Logger — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Logger 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
MyTestDataset Ran
ran on a synthesized input · constructed an object; no result was computed
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("a28c9c0199cb6d53")
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 an instance of its own class, MyTestDataset: the run constructed the object and did not call it, so no result was computed. 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
from PIL import Image
from torchvision import transforms

torch.manual_seed(0)

# Create small sample data
# For cifar10: data should be numpy arrays of shape (N, 32, 32, 3) with uint8 values
# For other datasets: data can be tensors
data_cifar10 = np.random.randint(0, 256, size=(2, 32, 32, 3), dtype=np.uint8)
data_other = torch.randn(2, 4, 4)

# Create labels
labels = torch.tensor([0, 1])

# Create a simple transform that works with PIL images
transform_cifar10 = transforms.Compose([
    transforms.ToTensor(),
])

# Create a simple transform that works with tensors
transform_other = transforms.Compose([
    transforms.ToTensor(),
])

# Test with cifar10 dataset
ARGS = {
    "data": data_cifar10,
    "labels": labels,
    "transform": transform_cifar10,
    "dataset": "cifar10"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type MyTestDataset — the run constructed an instance of its own class and did not call it; no shape 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
MyTrainDataset Ran
ran on a synthesized input · constructed an object; no result was computed
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("de641f731f9c0c61")
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 an instance of its own class, MyTrainDataset: the run constructed the object and did not call it, so no result was computed. 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
from PIL import Image
from torchvision import transforms

torch.manual_seed(0)

# Create small dummy image data (batch_size=2, height=4, width=3, channels=3)
data = np.random.randint(0, 256, (2, 4, 3, 3), dtype=np.uint8)

# Create small dummy labels
labels = np.array([0, 1])

# Create a simple transform that converts PIL image to tensor
transform = transforms.Compose([
    transforms.ToTensor(),
])

ARGS = {
    "data": data,
    "labels": labels,
    "transform": transform,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type MyTrainDataset — the run constructed an instance of its own class and did not call it; no shape 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
NtXentLoss Ran
ran on a synthesized input · constructed an object; no result was computed
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("b2b75636ac168c2e")
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 an instance of its own class, NtXentLoss: the run constructed the object and did not call it, so no result was computed. 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)

# NtXentLoss takes batch_size and temperature as constructor arguments
batch_size = 2
temperature = 0.07

ARGS = {
    'batch_size': batch_size,
    'temperature': temperature
}
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 NtXentLoss — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a NtXentLoss 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
calculate_hamming Ran
ran, violated the contract we drafted
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("60f0a788af0cca27")
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

# B1 is a vector of shape [n]
# B2 is a vector of shape [r*n], which will be treated as [r, n] when transposed
# The function computes: 0.5 * (q - np.dot(B1, B2.T))
# where q = B2.shape[1], so B2 should be 2D with shape [r, n]

n = 4  # dimension of vectors
r = 3  # number of vectors in B2

# B1: shape [n] = [4]
B1 = np.random.randn(n)

# B2: shape [r, n] = [3, 4]
# When we do B2.transpose(), it becomes [n, r] = [4, 3]
# np.dot(B1, B2.transpose()) will be [4] @ [4, 3] = [3]
# q = B2.shape[1] = 4
B2 = np.random.randn(r, n)

ARGS = {
    "B1": B1,
    "B2": B2
}
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
calculate_top_map Ran
ran; our contract draft was wrong, not the code
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("2713054f03946cac")
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, 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

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

# Parameters for calculate_top_map
# qB: {-1,+1}^{mxq} query bits - shape (num_query, num_bits)
num_query = 2
num_bits = 4
qB = np.random.choice([-1, 1], size=(num_query, num_bits)).astype(np.float32)

# rB: {-1,+1}^{nxq} retrieval bits - shape (num_retrieval, num_bits)
num_retrieval = 3
rB = np.random.choice([-1, 1], size=(num_retrieval, num_bits)).astype(np.float32)

# queryL: {0,1}^{mxl} query labels - shape (num_query, num_labels)
num_labels = 5
queryL = np.random.randint(0, 2, size=(num_query, num_labels)).astype(np.float32)

# retrievalL: {0,1}^{nxl} retrieval labels - shape (num_retrieval, num_labels)
retrievalL = np.random.randint(0, 2, size=(num_retrieval, num_labels)).astype(np.float32)

# topk: number of top results to consider
topk = 2

ARGS = {
    "qB": qB,
    "rB": rB,
    "queryL": queryL,
    "retrievalL": retrievalL,
    "topk": topk
}
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, 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
compress Ran
ran; our contract draft was wrong, not the code
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("9d8f8e1077379f49")
How it ran

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

import torch
import torch.nn as nn
import numpy as np
from torch.autograd import Variable

torch.manual_seed(0)

# Define a simple encoder model that outputs discrete codes
class DiscreteEncoder(nn.Module):
    def __init__(self, input_dim, code_dim):
        super(DiscreteEncoder, self).__init__()
        self.fc = nn.Linear(input_dim, code_dim)
    
    def forward(self, x):
        # Output discrete binary codes
        out = self.fc(x)
        return (out > 0).float()

# Create tiny train and test data loaders
# Each item is a (data, target) tuple
train_data = [
    (torch.randn(2, 4), torch.tensor([0, 1])),
    (torch.randn(2, 4), torch.tensor([1, 0])),
]

test_data = [
    (torch.randn(2, 4), torch.tensor([0, 0])),
    (torch.randn(2, 4), torch.tensor([1, 1])),
]

# Create the encoder model
encode_discrete = DiscreteEncoder(input_dim=4, code_dim=3)

# Set device to CPU
device = 'cpu'

ARGS = {
    'train': train_data,
    'test': test_data,
    'encode_discrete': encode_discrete,
    'device': device,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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_cifar Ran
ran on a synthesized input
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("dd5e14fd69dff667")
How it ran

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

import numpy as np
import tempfile
import os
from unittest.mock import Mock, patch

# Create mock CIFAR10 datasets
class MockCIFAR10:
    def __init__(self, train=True):
        # Create small mock data: 100 samples per class for train, 50 for test
        if train:
            self.data = np.random.randint(0, 256, (1000, 32, 32, 3), dtype=np.uint8)
            self.targets = np.tile(np.arange(10), 100)
        else:
            self.data = np.random.randint(0, 256, (500, 32, 32, 3), dtype=np.uint8)
            self.targets = np.tile(np.arange(10), 50)

# Mock the torchvision datasets module
def mock_cifar10_constructor(root='./data/cifar10/', train=True, download=False):
    return MockCIFAR10(train=train)

# Patch torchvision.datasets.CIFAR10
import torchvision.datasets as dsets
original_cifar10 = dsets.CIFAR10
dsets.CIFAR10 = mock_cifar10_constructor

# Now the function can be called without actual downloads
ARGS = {}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torchvision — 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
hash Ran
ran on a synthesized input · constructed an object; no result was computed
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("446c9e1a244aca95")
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 an instance of its own class, hash: the run constructed the object and did not call it, so no result was computed. 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
from torch.autograd import Function

torch.manual_seed(0)

class hash(Function):
    @staticmethod
    def forward(ctx, input):
        return torch.sign(input)

    @staticmethod
    def backward(ctx, grad_output):
        return grad_output

# Create a small tensor for the input parameter
input_tensor = torch.randn(2, 4)

ARGS = {
    "input": input_tensor
}
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 hash — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a hash 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
hash_layer Ran
ran; our contract draft was wrong, not the code
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("6dfd48e834be71b6")
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
from torch.autograd import Function

torch.manual_seed(0)

class hash(Function):

    @staticmethod
    def forward(ctx, input):
        return torch.sign(input)

    @staticmethod
    def backward(ctx, grad_output):
        return grad_output

def hash_layer(input):
    return hash.apply(input)

# Create a small tensor for testing
input_tensor = torch.randn(2, 4)

ARGS = {
    'input': input_tensor
}
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
Base_Model Not yet run
unverified (harvested, no recorded run)
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("7eebc68d4cc2574e")
How it ran

no invocation recorded.

CIBHash Not yet run
unverified (harvested, no recorded run)
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("b95d18b296ff8c21")
How it ran

no invocation recorded.

Data Not yet run
unverified (harvested, no recorded run)
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("6ae75be554ee6746")
How it ran

no invocation recorded.

LabeledData Not yet run
unverified (harvested, no recorded run)
qiuzx2/CIBHash/model/CIBHash.py
code served (permissive licence) · get_code("04adc0d5624cd788")
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, 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 15 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

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

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

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

Abstract

Many unsupervised hashing methods are implicitly established on the idea of reconstructing the input data, which basically encourages the hashing codes to retain as much information of original data as possible. However, this requirement may force the models spending lots of their effort on reconstructing the unuseful background information, while ignoring to preserve the discriminative semantic information that is more important for the hashing task. To tackle this problem, inspired by the recent success of contrastive learning in learning continuous representations, we propose to adapt this framework to learn binary hashing codes. Specifically, we first propose to modify the objective function to meet the specific requirement of hashing and then introduce a probabilistic binary representation layer into the model to facilitate end-to-end training of the entire model. We further prove the strong connection between the proposed contrastive-learning-based hashing method and the mutual information, and show that the proposed model can be considered under the broader framework of the information bottleneck (IB). Under this perspective, a more general hashing model is naturally obtained. Extensive experimental results on three benchmark image datasets demonstrate that the proposed hashing method significantly outperforms existing baselines.

For agents

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

get_harvested_code_for_paper("2105.06138")
get_code_for_paper("2105.06138")
have("2105.06138")

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

curl https://syntology.ai/api/ran/2105.06138.json

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

[![Syntology run record](https://syntology.ai/api/ran/2105.06138.svg)](https://syntology.ai/paper/2105.06138)

Connect an agent — have() is free.