SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2201.11732 · ICML · 2022

IGLUE: A Benchmark for Transfer Learning across Modalities, Tasks, and Languages

Ivan Vulić, Siva Reddy, Fangyu Liu, Emanuele Bugliarello, Jonas Pfeiffer, Desmond Elliott, Edoardo Ponti

arXiv · PDF · Open in the Atlas

Code that ran

17 of 21 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 2 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 7 failures of our instrument; 4 no recorded run); 1 of them was also run on the host, not in the jail, by the peer-agreement lane that set its level, 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/2201.11732.json · badge: /api/ran/2201.11732.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
e-bug/volta canonical (classifier label)
classifier basis not recorded on the link
15 ran · 4 no recorded run
e-bug/iglue canonical (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
FunctionStatusWhere it lives
swish Ran
ran; our contract draft was wrong, not the code
e-bug/volta/volta/encoders.py
code served (permissive licence) · get_code("0f786c407fb1ee4c")
How it ran

Ran in Syntology's jail on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) 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. The comparison behind this sample's level is a different run: the peer-agreement lane ran it again on the host, not in the jail (report dated 2026-09-04T22:33:52Z). If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function swish takes a single parameter 'x' which is used with:
# - torch.sigmoid(x) - requires a tensor
# - x * ... - requires a tensor that supports multiplication
# So x should be a tensor

x = torch.randn(2, 4)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
Embedding Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT) · constructed an object; no result was computed
e-bug/volta/volta/m3p_transformer.py
code served (permissive licence) · get_code("a3e27bd1e8b628ed")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned an instance of its own class, Embedding: 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)

num_embeddings = 10
embedding_dim = 4
padding_idx = 0

ARGS = {
    'num_embeddings': num_embeddings,
    'embedding_dim': embedding_dim,
    'padding_idx': padding_idx
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Embedding — 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 Embedding 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
Linear Ran
ran; our contract draft was wrong, not the code · constructed an object; no result was computed
e-bug/volta/volta/m3p_transformer.py
code served (permissive licence) · get_code("b5254f10aa821ab8")
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 an instance of its own class, Linear: 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)

ARGS = {
    "in_features": 4,
    "out_features": 3,
    "bias": True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Linear — 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 Linear 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
LoadLoss Ran
ran on a synthesized input
e-bug/volta/volta/task_utils.py
code served (permissive licence) · get_code("085e162f35b24d79")
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 function. 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)

def triplet_loss(rank_scores, target, margin=0.2):
    scores = torch.sigmoid(rank_scores)
    pos = scores[:, :1]
    neg = scores[:, 1:]
    rank_loss = torch.clamp(margin + neg - pos, 0)
    return rank_loss.mean()

LossMap = {'BCEWithLogitLoss': nn.BCEWithLogitsLoss(reduction='mean'), 'CrossEntropyLoss': nn.CrossEntropyLoss(), 'TripletLoss': triplet_loss}

def LoadLoss(args, task_cfg, task_id):
    task = 'TASK' + task_id
    loss_name = args.loss or task_cfg[task]['loss']
    loss = LossMap[loss_name]
    return loss

class Args:
    def __init__(self, loss=None):
        self.loss = loss

task_cfg = {
    'TASK001': {'loss': 'CrossEntropyLoss'},
    'TASK002': {'loss': 'BCEWithLogitLoss'},
    'TASK003': {'loss': 'TripletLoss'}
}

args = Args(loss='TripletLoss')
task_id = '001'

ARGS = {
    'args': args,
    'task_cfg': task_cfg,
    'task_id': task_id
}
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 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
collate_fn Ran
ran on a synthesized input
e-bug/iglue/features_extraction/detectron2_proposal_maxnms.py
code served (permissive licence) · get_code("4c2f89511d326a22")
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 torch
import numpy as np

torch.manual_seed(0)

# The collate_fn expects a batch which is an iterable of dicts
# Each dict should have 'img_id' and 'img' keys
batch = [
    {
        'img_id': 'img_001',
        'img': torch.randn(3, 4, 4)
    },
    {
        'img_id': 'img_002',
        'img': torch.randn(3, 4, 4)
    }
]

ARGS = {
    'batch': batch
}
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 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
coordinate_embeddings Ran
ran on a synthesized input
e-bug/volta/volta/embeddings.py
code served (permissive licence) · get_code("58c8b6cbad44801a")
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 4, dtype family float, shape (2, 3, 4, 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 torch

torch.manual_seed(0)

# boxes: [BS, K, 4] where BS is batch size, K is number of boxes, 4 is [x1, y1, x2, y2]
# Create a small batch with 2 samples, 3 boxes per sample, 4 coordinates each
boxes = torch.tensor([
    [[10.0, 20.0, 50.0, 80.0],
     [15.0, 25.0, 60.0, 85.0],
     [5.0, 10.0, 40.0, 70.0]],
    [[12.0, 22.0, 52.0, 82.0],
     [18.0, 28.0, 62.0, 88.0],
     [8.0, 12.0, 42.0, 72.0]]
])

# dim: embedding dimension for sin/cos
dim = 4

ARGS = {
    "boxes": boxes,
    "dim": dim
}
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 4, dtype family float, shape (2, 3, 4, 8) — shape (2, 3, 4, 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
create_position_ids_from_input_ids Ran
ran; our contract draft was wrong, not the code
e-bug/volta/volta/embeddings.py
pointer only (licence: MIT); the stored copy get_code serves, from Gabriel819/zeroshot_ner, is pointer only (licence: NONE) · get_code("d4903874fdec3763")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from Gabriel819/zeroshot_ner, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

filename_to_url Ran
ran; our contract draft was wrong, not the code
e-bug/volta/volta/utils.py
code served (permissive licence) · get_code("82874b46e206faf3")
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 os
import sys
import json
import tempfile
from pathlib import Path

# Create a temporary directory to serve as cache
temp_dir = tempfile.mkdtemp()

# Create a test file and its metadata file
test_filename = "model.bin"
test_file_path = os.path.join(temp_dir, test_filename)
meta_file_path = test_file_path + ".json"

# Write a dummy model file
with open(test_file_path, 'w') as f:
    f.write("dummy model content")

# Write metadata file with url and etag
metadata = {
    "url": "https://example.com/model.bin",
    "etag": "abc123def456"
}
with open(meta_file_path, 'w') as f:
    json.dump(metadata, f)

ARGS = {
    "filename": test_filename,
    "cache_dir": temp_dir
}
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 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
gelu Ran
ran, honoured the contract we drafted
e-bug/volta/volta/encoders.py
pointer only (licence: MIT); the stored copy get_code serves, from tanmay1618/text_to_sql_bert, is pointer only (licence: NONE) · get_code("fdc64f4c72036ae4")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from tanmay1618/text_to_sql_bert, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

gelu Ran
ran, honoured the contract we drafted
e-bug/volta/volta/m3p_transformer.py
code served (permissive licence) · get_code("66c21a3911091d29")
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 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 math
import torch

torch.manual_seed(0)

# The gelu function takes a single parameter 'x' which is used in:
# - 0.5 * x (scalar multiplication)
# - torch.erf(x / math.sqrt(2.0)) (error function)
# So x should be a tensor that supports these operations

x = torch.randn(2, 4)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
iou Ran
ran on a synthesized input
e-bug/volta/volta/datasets/concept_cap_dataset.py
pointer only (licence: MIT); the stored copy get_code serves, from facebookresearch/vilbert-multi-task, is pointer only (licence: MIT) · get_code("9499ed1255d97132")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from facebookresearch/vilbert-multi-task, and MIT, as GitHub detects it on that repository, is in the permissive set, but a licence decision keeps the body a pointer: its licence file names an outside party, so it may cover vendored code rather than the repository owner's files (recorded as pointer:licence_names_third_party_corporation). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

kl_1601 Ran
ran on a synthesized input
e-bug/volta/volta/losses.py
code served (permissive licence) · get_code("a2faafba33358747")
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 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Build arguments for kl_1601 function
# prediction_scores_v: tensor of shape (batch, seq_len, num_classes) used with F.log_softmax(dim=2)
prediction_scores_v = torch.randn(2, 3, 4)

# weight: scalar float for weighting the loss
weight = 0.5

# label: tensor of shape (batch, seq_len) with 0/1 values, used with (label == 1)
label = torch.tensor([[1, 0, 1], [0, 1, 0]], dtype=torch.long)

# image_cls: target probabilities for KLDivLoss, shape (batch, seq_len, num_classes)
# Must match prediction_scores_v shape for KLDivLoss
image_cls = torch.softmax(torch.randn(2, 3, 4), dim=2)

# obj_labels, obj_confs, attr_labels, attr_confs: not used in function body
obj_labels = None
obj_confs = None
attr_labels = None
attr_confs = None

# image_feat: not used in function body
image_feat = None

ARGS = {
    'prediction_scores_v': prediction_scores_v,
    'weight': weight,
    'label': label,
    'image_cls': image_cls,
    'image_feat': image_feat,
    'obj_labels': obj_labels,
    'obj_confs': obj_confs,
    'attr_labels': attr_labels,
    'attr_confs': attr_confs,
}
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 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
load_obj_tsv Ran
ran on a synthesized input
e-bug/iglue/features_extraction/tsv_to_h5.py
code served (permissive licence) · get_code("fd48f6a6c3aa0b29")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family 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 csv
import base64
import tempfile
import numpy as np

# Create a temporary TSV file with valid object detection data
def create_test_tsv():
    # Create a temporary file
    temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.tsv')
    
    # Write header
    fieldnames = ['img_id', 'img_h', 'img_w', 'objects_id', 'objects_conf', 'attrs_id', 'attrs_conf', 'num_boxes', 'boxes', 'features']
    writer = csv.DictWriter(temp_file, fieldnames=fieldnames, delimiter='\t')
    
    # Create sample data for 2 images
    for img_idx in range(2):
        num_boxes = 3
        feature_dim = 2048
        
        # Generate random arrays
        objects_id = np.random.randint(0, 1600, size=(num_boxes,), dtype=np.int64)
        objects_conf = np.random.rand(num_boxes).astype(np.float32)
        attrs_id = np.random.randint(0, 400, size=(num_boxes,), dtype=np.int64)
        attrs_conf = np.random.rand(num_boxes).astype(np.float32)
        boxes = np.random.rand(num_boxes, 4).astype(np.float32) * 100
        features = np.random.rand(num_boxes, feature_dim).astype(np.float32)
        
        # Encode arrays to base64
        row = {
            'img_id': f'image_{img_idx}',
            'img_h': 480,
            'img_w': 640,
            'objects_id': base64.b64encode(objects_id.tobytes()).decode('utf-8'),
            'objects_conf': base64.b64encode(objects_conf.tobytes()).decode('utf-8'),
            'attrs_id': base64.b64encode(attrs_id.tobytes()).decode('utf-8'),
            'attrs_conf': base64.b64encode(attrs_conf.tobytes()).decode('utf-8'),
            'num_boxes': num_boxes,
            'boxes': base64.b64encode(boxes.tobytes()).decode('utf-8'),
            'features': base64.b64encode(features.tobytes()).decode('utf-8'),
        }
        writer.writerow(row)
    
    temp_file.close()
    return temp_file.name

tsv_file_path = create_test_tsv()

ARGS = {
    'fname': tsv_file_path,
    'topk': None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
mse_2048 Ran
ran on a synthesized input
e-bug/volta/volta/losses.py
code served (permissive licence) · get_code("1adf127958517c64")
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 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Build small tensors with distinct dimensions
batch_size = 2
feat_dim = 4
num_classes = 3

# prediction_scores_v: model output, shape (batch, num_classes, feat_dim) or similar
# Based on MSELoss usage, it should match image_target shape
prediction_scores_v = torch.randn(batch_size, num_classes, feat_dim)

# weight: scalar controlling loss contribution
weight = 1.0

# label: shape (batch, num_classes) with values 0 or 1
label = torch.tensor([[1, 0, 1], [0, 1, 0]], dtype=torch.long)

# image_cls: not used in function body, can be None or dummy
image_cls = None

# image_feat: target features, shape (batch, num_classes, feat_dim)
image_feat = torch.randn(batch_size, num_classes, feat_dim)

# obj_labels: not used in function body
obj_labels = None

# obj_confs: not used in function body
obj_confs = None

# attr_labels: not used in function body
attr_labels = None

# attr_confs: not used in function body
attr_confs = None

ARGS = {
    'prediction_scores_v': prediction_scores_v,
    'weight': weight,
    'label': label,
    'image_cls': image_cls,
    'image_feat': image_feat,
    'obj_labels': obj_labels,
    'obj_confs': obj_confs,
    'attr_labels': attr_labels,
    'attr_confs': attr_confs,
}
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 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
resume Ran
ran; our contract draft was wrong, not the code
e-bug/volta/volta/train_utils.py
code served (permissive licence) · get_code("64fe66e56c74b1f6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 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 torch
import torch.nn as nn
import torch.optim as optim
import tempfile
import os

torch.manual_seed(0)

# Create a simple model
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)

model = SimpleModel()

# Create optimizer
optimizer = optim.Adam(model.parameters(), lr=0.001)

# Create scheduler
scheduler = optim.lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.1)

# Create a simple tb_logger stand-in (can be any object since it's just stored/retrieved)
tb_logger = {"log_dir": "/tmp/logs", "step": 0}

# Create a temporary checkpoint file
temp_dir = tempfile.mkdtemp()
checkpoint_path = os.path.join(temp_dir, "checkpoint.pth")

# Save a checkpoint to test the resume function
checkpoint = {
    'model_state_dict': model.state_dict(),
    'optimizer_state_dict': optimizer.state_dict(),
    'scheduler_state_dict': scheduler.state_dict(),
    'global_step': 100,
    'epoch_id': 5,
    'tb_logger': tb_logger,
    'score': 0.85
}
torch.save(checkpoint, checkpoint_path)

ARGS = {
    'path': checkpoint_path,
    'model': model,
    'optimizer': optimizer,
    'scheduler': scheduler,
    'tb_logger': tb_logger
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
torch — 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
triplet_loss Ran
ran on a synthesized input
e-bug/volta/volta/task_utils.py
code served (permissive licence) · get_code("b8c86cabecffbff9")
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 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 torch

torch.manual_seed(0)

# The function expects:
# - rank_scores: a tensor that will be passed through sigmoid, then indexed as [:, :1] and [:, 1:]
#   This means it needs shape (batch_size, num_scores) where num_scores >= 2
# - target: passed to the function but not actually used in the body
# - margin: a float (optional, defaults to 0.2)

batch_size = 2
num_scores = 4

rank_scores = torch.randn(batch_size, num_scores)
target = torch.randint(0, 2, (batch_size,))
margin = 0.2

ARGS = {
    'rank_scores': rank_scores,
    'target': target,
    'margin': margin,
}
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 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
url_to_filename Ran
ran; our contract draft was wrong, not the code
e-bug/volta/volta/utils.py
code served (permissive licence) · get_code("5eff22fa0a651276")
How it ran

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

from hashlib import sha256

ARGS = {
    "url": "https://example.com/model/checkpoint.bin",
    "etag": "abc123def456"
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
ForwardModelsVal Not yet run
unverified (harvested, no recorded run)
e-bug/volta/volta/task_utils.py
code served (permissive licence) · get_code("07cb4167108cab0b")
How it ran

no invocation recorded.

cached_path Not yet run
unverified (harvested, no recorded run)
e-bug/volta/volta/utils.py
code served (permissive licence) · get_code("1873b62b661f9fda")
How it ran

no invocation recorded.

load_tf_weights_in_bert Not yet run
unverified (harvested, no recorded run)
e-bug/volta/volta/encoders.py
code served (permissive licence) · get_code("baa5766f4566aafc")
How it ran

no invocation recorded.

nce_2048 Not yet run
unverified (harvested, no recorded run)
e-bug/volta/volta/losses.py
code served (permissive licence) · get_code("60cab169db54e986")
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, 16 are also present in other papers whose code we hold (15 with at least one paper dated earlier, 4 the same month, 10 later) and 5 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
swish 206 other papers
90 dated earlier, 1 the same month, 111 later, 4 same year, order not determinable · 44 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1805.12114 (2018-05)
Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models
1809.01357 (2018-09)
Zero Shot Learning for Code Education: Rubric Sampling with Deep Learning Inference
and 203 more not named here
gelu 198 other papers
96 dated earlier, 1 the same month, 97 later, 4 same year, order not determinable · 32 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1901.10125 (2019-01)
Glyce: Glyph-vectors for Chinese Character Representations
and 195 more not named here
url_to_filename 48 other papers
32 dated earlier, 15 later, 1 same year, order not determinable · 8 only as an assembly brick
1610.02424 (2016-10)
Diverse Beam Search: Decoding Diverse Solutions from Neural Sequence Models
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
and 45 more not named here
create_position_ids_from_input_ids 28 other papers
2 dated earlier, 26 later
2107.00910 (2021-07)
Learned Token Pruning for Transformers
2107.14035 (2021-07)
ProtoTransformer: A Meta-Learning Approach to Providing Student Feedback
2202.07962 (2022-02)
Revisiting Parameter-Efficient Tuning: Are We Really There Yet?
and 25 more not named here
cached_path 27 other papers
22 dated earlier, 5 later · 5 only as an assembly brick
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1901.02860 (2019-01)
Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context
1908.02265 (2019-08)
ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks
and 24 more not named here
Embedding 18 other papers
6 dated earlier, 1 the same month, 11 later · 3 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
1905.12790 (2019-05)
A Generalized Framework of Sequence Generation with Application to Undirected Sequence Models
1911.02116 (2019-11)
Unsupervised Cross-lingual Representation Learning at Scale
and 15 more not named here
filename_to_url 18 other papers
13 dated earlier, 5 later
1908.09355 (2019-08)
Patient Knowledge Distillation for BERT Model Compression
1909.02209 (2019-09)
Semantics-aware BERT for Language Understanding
1911.02896 (2019-11)
Contextualized Sparse Representations for Real-Time Open-Domain Question Answering
and 15 more not named here
gelu 17 other papers
7 dated earlier, 1 the same month, 9 later
1905.12790 (2019-05)
A Generalized Framework of Sequence Generation with Application to Undirected Sequence Models
1906.08237 (2019-06)
XLNet: Generalized Autoregressive Pretraining for Language Understanding
1909.05311 (2019-09)
Graph-Based Reasoning over Heterogeneous External Knowledge for Commonsense Question Answering
and 14 more not named here
load_tf_weights_in_bert 12 other papers
10 dated earlier, 2 later · 1 only as an assembly brick
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1906.01698 (2019-06)
Open Sesame: Getting Inside BERT's Linguistic Knowledge
1908.02265 (2019-08)
ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks
and 9 more not named here
Linear 2 other papers
2 dated earlier
1905.12790 (2019-05)
A Generalized Framework of Sequence Generation with Application to Undirected Sequence Models
1911.02116 (2019-11)
Unsupervised Cross-lingual Representation Learning at Scale
kl_1601 2 other papers
2 dated earlier
2109.04448 (2021-09)
Vision-and-Language or Vision-for-Language? On Cross-Modal Influence in Multimodal Transformers
2109.13238 (2021-09)
Visually Grounded Reasoning across Languages and Cultures
mse_2048 2 other papers
2 dated earlier
2109.04448 (2021-09)
Vision-and-Language or Vision-for-Language? On Cross-Modal Influence in Multimodal Transformers
2109.13238 (2021-09)
Visually Grounded Reasoning across Languages and Cultures
nce_2048 2 other papers
2 dated earlier
2109.04448 (2021-09)
Vision-and-Language or Vision-for-Language? On Cross-Modal Influence in Multimodal Transformers
2109.13238 (2021-09)
Visually Grounded Reasoning across Languages and Cultures
collate_fn 1 other paper
1 dated earlier
2109.13238 (2021-09)
Visually Grounded Reasoning across Languages and Cultures
iou 1 other paper
1 dated earlier
1912.02315 (2019-12)
12-in-1: Multi-Task Vision and Language Representation Learning
resume 1 other paper
1 later
2203.15867 (2022-03)
Image Retrieval from Contextual Descriptions

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

Reliable evaluation benchmarks designed for replicability and comprehensiveness have driven progress in machine learning. Due to the lack of a multilingual benchmark, however, vision-andlanguage research has mostly focused on English language tasks. To fill this gap, we introduce the Image-Grounded Language Understanding Evaluation benchmark. IGLUE brings togetherby both aggregating pre-existing datasets and creating new ones-visual question answering, cross-modal retrieval, grounded reasoning, and grounded entailment tasks across 20 diverse languages. Our benchmark enables the evaluation of multilingual multimodal models for transfer learning, not only in a zero-shot setting, but also in newly defined few-shot learning setups. Based on the evaluation of the available state-of-the-art models, we find that translate-test transfer is superior to zero-shot transfer and that few-shot learning is hard to harness for many tasks. Moreover, downstream performance is partially explained by the amount of available unlabelled textual data for pretraining, and only weakly by the typological distance of target-source languages. We hope to encourage future research efforts in this area by releasing the benchmark to the community.

For agents

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

get_harvested_code_for_paper("2201.11732")
get_code_for_paper("2201.11732")
have("2201.11732")

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

curl https://syntology.ai/api/ran/2201.11732.json

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

[![Syntology run record](https://syntology.ai/api/ran/2201.11732.svg)](https://syntology.ai/paper/2201.11732)

Connect an agent — have() is free.