SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2204.08790 · NeurIPS · 2022

ELEVATER: A Benchmark and Toolkit for Evaluating Language-Augmented Visual Models

Chunyuan Li, Jianfeng Gao, Jyoti Aneja, Zicheng Liu, -Madison, Jianwei Yang, Pengchuan Zhang, Haotian Liu, Liunian Li, Ping Jin, Houdong Hu, Yong Lee, and 1 more

arXiv · PDF · Open in the Atlas

Code that ran

12 of 20 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 4 failures of our instrument; 8 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/2204.08790.json · badge: /api/ran/2204.08790.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
eric-ai-lab/pevit canonical (classifier label)
classifier basis not recorded on the link
7 ran · 2 no recorded run
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC canonical (classifier label)
classifier basis not recorded on the link
5 ran · 6 no recorded run
FunctionStatusWhere it lives
window_partition Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/models/cls_swin.py
code served (permissive licence) · get_code("144d10b49baeb8a6")
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 4, dtype family float, shape (8, 4, 4, 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 2026-09-04T22:38:58Z). If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a tensor with shape (B, H, W, C)
# B=2 (batch), H=8 (height), W=8 (width), C=4 (channels)
# Using distinct sizes: 2, 8, 8, 4
x = torch.randn(2, 8, 8, 4)

# window_size must divide both H and W evenly
# H=8, W=8, so window_size=4 works (8 // 4 = 2)
window_size = 4

ARGS = {
    "x": x,
    "window_size": window_size,
}
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 4, dtype family float, shape (8, 4, 4, 4) — shape (8, 4, 4, 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
extract_finetune_results Ran
ran on a synthesized input
eric-ai-lab/pevit/read_results.py
code served (permissive licence) · get_code("0fcad17cd3cdfb8d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (1, 1) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import os
import tempfile
import shutil

# Create a temporary directory structure with mock log files
temp_dir = tempfile.mkdtemp()
dataset_name = 'test_dataset'
dataset_path = os.path.join(temp_dir, dataset_name)
os.makedirs(dataset_path, exist_ok=True)

# Create mock log files with finetuning results
num_samples = '5'
file_name = f'finetuning_{num_samples}_0.txt'
file_path = os.path.join(dataset_path, file_name)

# Write mock log content with accuracy and trainable params
mock_log_content = """epoch 1 loss 0.5
epoch 2 loss 0.3
trainable params: 2.5M
final accuracy: 85.5%"""

with open(file_path, 'w') as f:
    f.write(mock_log_content)

# Create additional mock files for different random seeds
for seed in range(1, 3):
    file_name = f'finetuning_{num_samples}_{seed}.txt'
    file_path = os.path.join(dataset_path, file_name)
    mock_log_content = f"""epoch 1 loss 0.5
epoch 2 loss 0.3
trainable params: 2.5M
final accuracy: {85.0 + seed}%"""
    with open(file_path, 'w') as f:
        f.write(mock_log_content)

ARGS = {
    'proj_path': temp_dir,
    'dataset_name': dataset_name,
    'num_samples_per_class': ['5'],
    'rs': 0
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (1, 1) — shape (1, 1) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
extract_finetune_results Ran
ran on a synthesized input
eric-ai-lab/pevit/read_txt.py
code served (permissive licence) · get_code("d276b7d04501c57c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (1, 1) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import os
import tempfile
import shutil

# Create a temporary directory structure to simulate the log files
temp_dir = tempfile.mkdtemp()
predictions_dir = os.path.join(temp_dir, 'predictions', 'finetuning_5')
os.makedirs(predictions_dir, exist_ok=True)

# Create sample log files with the expected format
sample_file_content = """Some training logs
best acc is: 92.5, other_metric: value
n_trainable_params is: 1024, config: data
final result, accuracy: 92.5%"""

# Create a file matching the dataset_name pattern
dataset_name = 'imagenet'
file_name = f'finetuning_5_{dataset_name}_results.txt'
file_path = os.path.join(predictions_dir, file_name)
with open(file_path, 'w') as f:
    f.write(sample_file_content)

# Define arguments for extract_finetune_results
ARGS = {
    'proj_path': temp_dir,
    'dataset_name': 'imagenet',
    'num_samples_per_class': ['5'],
    'rs': 42
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (1, 1) — shape (1, 1) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
extract_finetune_results_para Ran
ran on a synthesized input
eric-ai-lab/pevit/read_txt.py
code served (permissive licence) · get_code("ced21f6a4b3113c2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (1, 1) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import os
import tempfile
import shutil

# Create a temporary directory structure for the test
temp_dir = tempfile.mkdtemp()
log_path = temp_dir

# Create the expected directory structure
predictions_dir = os.path.join(log_path, 'predictions', 'finetuning_5')
os.makedirs(predictions_dir, exist_ok=True)

# Create sample txt files with the expected format
sample_file_content_1 = """some training data
n_trainable_params is: 12345, other_info
best acc is: 0.875, final_metric
result, metric: 92.5%"""

sample_file_content_2 = """training iteration 1
n_trainable_params is: 67890, other_info
best acc is: 0.920, final_metric
result, metric: 95.0%"""

# Write sample files
with open(os.path.join(predictions_dir, 'finetuning_5_imagenet_test.txt'), 'w') as f:
    f.write(sample_file_content_1)

with open(os.path.join(predictions_dir, 'finetuning_5_cifar_test.txt'), 'w') as f:
    f.write(sample_file_content_2)

# Define the ARGS dictionary with all parameters
ARGS = {
    'proj_path': log_path,
    'dataset_name': 'imagenet',
    'num_samples_per_class': ['5'],
    'rs': 42
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (1, 1) — shape (1, 1) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_cls_model Ran
ran on a synthesized input
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/models/cls_example.py
code served (permissive licence) · get_code("aa72222fcaba5411")
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 Example. 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 import nn

torch.manual_seed(0)

class Example(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)
    
    def forward_features(self):
        """
        This method is called to extract features for evaluation.
        """
        pass

class SimpleConfig:
    def __init__(self):
        self.model_name = "example"
        self.hidden_size = 64
        self.num_classes = 10

config = SimpleConfig()

ARGS = {
    "config": config
}
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 Example — a Example output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Example 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
get_model Ran
ran on a synthesized input
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/models/mae.py
code served (permissive licence) · get_code("853c0a124b0d326e")
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 VisionTransformer. 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
from functools import partial
from unittest.mock import MagicMock

torch.manual_seed(0)

class MockConfig:
    def __init__(self):
        self.MODEL = MagicMock()
        self.TEST = MagicMock()
        
        spec = MagicMock()
        spec.PATCH_SIZE = 16
        spec.EMBED_DIM = 64
        spec.DEPTH = 2
        spec.NUM_HEADS = 4
        spec.MLP_RATIO = 4.0
        spec.QKV_BIAS = True
        spec.GLOBAL_POOL = False
        
        self.MODEL.SPEC = spec
        self.TEST.MODEL_FILE = "/tmp/model.pt"

def create_mock_checkpoint():
    """Create a mock checkpoint with model state dict"""
    state_dict = {
        'patch_embed.proj.weight': torch.randn(64, 3, 16, 16),
        'patch_embed.proj.bias': torch.randn(64),
        'cls_token': torch.randn(1, 1, 64),
        'pos_embed': torch.randn(1, 197, 64),
        'pos_drop': torch.nn.Identity(),
        'norm.weight': torch.randn(64),
        'norm.bias': torch.randn(64),
        'blocks.0.norm1.weight': torch.randn(64),
        'blocks.0.norm1.bias': torch.randn(64),
        'blocks.1.norm1.weight': torch.randn(64),
        'blocks.1.norm1.bias': torch.randn(64),
    }
    return {'model': state_dict}

config = MockConfig()

# Mock torch.load to return our checkpoint
original_torch_load = torch.load
def mock_torch_load(path, **kwargs):
    return create_mock_checkpoint()

torch.load = mock_torch_load

ARGS = {
    'config': config
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
timm, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type VisionTransformer — a VisionTransformer output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a VisionTransformer 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
get_zeroshot_model Ran
ran on a synthesized input
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/models/clip_example.py
code served (permissive licence) · get_code("deeacd37b9b6b8a5")
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 Example. 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 import nn

torch.manual_seed(0)

class Config:
    def __init__(self):
        self.model_name = "example_model"
        self.hidden_dim = 4
        self.num_classes = 2

config = Config()

ARGS = {
    "config": config
}
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 Example — a Example output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Example 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
glorot_uniform Ran
ran; our contract draft was wrong, not the code
eric-ai-lab/pevit/vision_benchmark/evaluation/compacter_model.py
pointer only (licence: MIT); the stored copy get_code serves, from rabeehk/compacter, is pointer only (licence: NONE) · get_code("b2ee26e229353a37")
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 rabeehk/compacter, 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.

kronecker_product_einsum_batched Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
eric-ai-lab/pevit/vision_benchmark/evaluation/compacter_model.py
pointer only (licence: MIT); the stored copy get_code serves, from jkooy/Parameter-efficient-Fine-tuning-for-Vision-Transformers, is pointer only (licence: NONE) · get_code("59e2991d3f405758")
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 jkooy/Parameter-efficient-Fine-tuning-for-Vision-Transformers, 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.

read_json Ran
ran on a synthesized input
eric-ai-lab/pevit/read_results.py
code served (permissive licence) · get_code("4e6ec58f35d0977b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import tempfile
import shutil

# Create a temporary directory structure with test files
temp_dir = tempfile.mkdtemp()
dataset_subdir = os.path.join(temp_dir, 'test_dataset')
os.makedirs(dataset_subdir, exist_ok=True)

# Create test log files with the expected format
test_file_1 = os.path.join(dataset_subdir, 'finetuning_log_1.txt')
with open(test_file_1, 'w') as f:
    f.write("epoch 1 loss: 0.5\n")
    f.write("epoch 2 loss: 0.3\n")
    f.write("trainable params: 12.5M accuracy: 95.5%\n")

test_file_2 = os.path.join(dataset_subdir, 'finetuning_log_2.txt')
with open(test_file_2, 'w') as f:
    f.write("epoch 1 loss: 0.4\n")
    f.write("trainable params: 8.3M accuracy: 92.3%\n")

test_file_3 = os.path.join(dataset_subdir, 'finetuning_log_3.txt')
with open(test_file_3, 'w') as f:
    f.write("epoch 1 loss: 0.6\n")
    f.write("trainable params: 15.7M accuracy: 88.9%\n")

ARGS = {
    'log_path': temp_dir,
    'dataset_name': 'test_dataset',
    'file_prefix': 'finetuning_'
}
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 2, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
read_json Ran
ran on a synthesized input
eric-ai-lab/pevit/read_txt.py
code served (permissive licence) · get_code("5f28aa08fa3d16f1")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import tempfile
import shutil

# Create a temporary directory structure for testing
temp_dir = tempfile.mkdtemp()
log_dir = os.path.join(temp_dir, 'predictions', 'finetuning_5')
os.makedirs(log_dir, exist_ok=True)

# Create sample txt files with the expected format
sample_content_1 = """
Training on dataset_mnist
n_trainable_params is: 12345, other_data
best acc is: 95.67, other_info
loss: 0.5, accuracy:92.3%
"""

sample_content_2 = """
Training on dataset_cifar
n_trainable_params is: 67890, other_data
best acc is: 88.42, other_info
loss: 0.3, accuracy:85.5%
"""

with open(os.path.join(log_dir, 'results_dataset_mnist.txt'), 'w') as f:
    f.write(sample_content_1)

with open(os.path.join(log_dir, 'results_dataset_cifar.txt'), 'w') as f:
    f.write(sample_content_2)

ARGS = {
    'log_path': temp_dir,
    'dataset_name': 'dataset_mnist',
    'file_prefix': ''
}
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 2, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
window_reverse Ran
ran; our contract draft was wrong, not the code
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/models/cls_swin.py
code served (permissive licence) · get_code("61bf152e6a42a184")
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 4, dtype family float, shape (2, 8, 12, 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 torch

torch.manual_seed(0)

# Parameters for window_reverse
# windows: (num_windows*B, window_size, window_size, C)
# The function expects windows to be partitioned from an image of size (B, H, W, C)
# where the image is divided into non-overlapping windows of size window_size x window_size

B = 2  # batch size
H = 8  # image height
W = 12  # image width
window_size = 4  # window size
C = 3  # number of channels

# Calculate number of windows
num_windows_h = H // window_size  # 2
num_windows_w = W // window_size  # 3
num_windows = num_windows_h * num_windows_w  # 6

# windows shape: (num_windows*B, window_size, window_size, C)
# = (6*2, 4, 4, 3) = (12, 4, 4, 3)
windows = torch.randn(num_windows * B, window_size, window_size, C)

ARGS = {
    "windows": windows,
    "window_size": window_size,
    "H": H,
    "W": W,
}
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 4, dtype family float, shape (2, 8, 12, 3) — shape (2, 8, 12, 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
ask_gpt3 Not yet run
unverified (harvested, no recorded run)
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/commands/extract_gpt3_knowledge.py
code served (permissive licence) · get_code("ef2f936b8e5c816f")
How it ran

no invocation recorded.

build_adapter_model Not yet run
unverified (harvested, no recorded run)
eric-ai-lab/pevit/vision_benchmark/evaluation/adapter_model.py
code served (permissive licence) · get_code("497088982a2f6f81")
How it ran

no invocation recorded.

build_compacter_model Not yet run
unverified (harvested, no recorded run)
eric-ai-lab/pevit/vision_benchmark/evaluation/compacter_model.py
code served (permissive licence) · get_code("052dba32d0be4f7e")
How it ran

no invocation recorded.

clone_loader Not yet run
unverified (harvested, no recorded run)
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/evaluation/full_model_finetune.py
code served (permissive licence) · get_code("ea0010b94502a91c")
How it ran

no invocation recorded.

get_cls_model Not yet run
unverified (harvested, no recorded run)
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/models/cls_swin.py
code served (permissive licence) · get_code("68f56d4679dbb36f")
How it ran

no invocation recorded.

get_model Not yet run
unverified (harvested, no recorded run)
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/models/declip.py
code served (permissive licence) · get_code("aead1abefc725217")
How it ran

no invocation recorded.

get_model Not yet run
unverified (harvested, no recorded run)
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/models/mocov3.py
code served (permissive licence) · get_code("c6f5f3a48a3d847d")
How it ran

no invocation recorded.

random_query Not yet run
unverified (harvested, no recorded run)
Computer-Vision-in-the-Wild/Elevater_Toolkit_IC/vision_benchmark/commands/extract_gpt3_knowledge.py
code served (permissive licence) · get_code("c35a97c3423834d4")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 20 distinct functions with an identity, 4 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 2 the same month, 2 later) and 16 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
window_reverse 227 other papers
45 dated earlier, 7 the same month, 173 later, 2 same year, order not determinable · 34 only as an assembly brick
1608.03983 (2016-08)
SGDR: Stochastic Gradient Descent with Warm Restarts
1711.10925 (2017-11)
Deep Image Prior
2003.13678 (2020-03)
Designing Network Design Spaces
and 224 more not named here
window_partition 223 other papers
42 dated earlier, 6 the same month, 173 later, 2 same year, order not determinable · 32 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
1711.10925 (2017-11)
Deep Image Prior
2003.13678 (2020-03)
Designing Network Design Spaces
and 220 more not named here
kronecker_product_einsum_batched 2 other papers
2 dated earlier
2106.04647 (2021-06)
Efficient Low-Rank Hypercomplex Adapter Layers
2203.16329 (2022-03)
Parameter-efficient Model Adaptation for Vision Transformers
glorot_uniform 1 other paper
1 dated earlier
2106.04647 (2021-06)
Efficient Low-Rank Hypercomplex Adapter Layers

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

Learning visual representations from natural language supervision has recently shown great promise in a number of pioneering works. In general, these language-augmented visual models demonstrate strong transferability to a variety of datasets and tasks. However, it remains challenging to evaluate the transferablity of these models due to the lack of easy-to-use evaluation toolkits and public benchmarks. To tackle this, we build ELEVATER 1 , the first benchmark and toolkit for evaluating (pre-trained) language-augmented visual models. ELEVATER is composed of three components. (i) Datasets. As downstream evaluation suites, it consists of 20 image classification datasets and 35 object detection datasets, each of which is augmented with external knowledge. (ii) Toolkit. An automatic hyper-parameter tuning toolkit is developed to facilitate model evaluation on downstream tasks. (iii) Metrics. A variety of evaluation metrics are used to measure sample-efficiency (zero-shot and few-shot) and parameter-efficiency (linear probing and full model fine-tuning). ELEVATER is platform for Computer Vision in the Wild (CVinW), and is publicly released at https://computer-vision-in-the-wild.github.io/ELEVATER/.

For agents

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

get_harvested_code_for_paper("2204.08790")
get_code_for_paper("2204.08790")
have("2204.08790")

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

curl https://syntology.ai/api/ran/2204.08790.json

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

[![Syntology run record](https://syntology.ai/api/ran/2204.08790.svg)](https://syntology.ai/paper/2204.08790)

Connect an agent — have() is free.