SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2004.06165 · 2020

Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks

Yejin Choi, Chunyuan Li, Jianfeng Gao, Lei Zhang, Lijuan Wang, Li, Pengchuan Zhang, Houdong Hu, Xiaowei Hu, Xiujun Li, Xi Yin, ♥ Li, and 1 more

arXiv · PDF · Open in the Atlas

Code that ran

13 of 23 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; 1 honoured the contract we drafted, 0 violated it and 9 have no contract verdict; 3 failures of our instrument; 10 no recorded run), measured 2026-11-21; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2004.06165.json · badge: /api/ran/2004.06165.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
microsoft/Oscar canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
milvlg/rosita unrelated (classifier label)
classifier basis: classifier_2025
9 ran · 7 no recorded run
ThanThoai/Visual-Question-Answering_Vietnamese community (classifier label)
classifier basis not recorded on the link
3 ran · 3 no recorded run
FunctionStatusWhere it lives
Backbone Ran
ran on a synthesized input · constructed an object; no result was computed
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("bb6ea058ef7c09e1")
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, Backbone: 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
import math

torch.manual_seed(0)

# Create a minimal config class with required attributes
class Config:
    def __init__(self):
        self.DROPOUT_R = 0.1

# Create model_dict with required keys
model_dict = {
    'HSIZE': 8,      # hidden size
    'HHEAD': 2,      # number of attention heads
    'HBASE': 4,      # base hidden size per head (HSIZE / HHEAD)
    'HFF': 16,       # feed-forward hidden size
    'LAYER': 2       # number of layers
}

# Verify HSIZE = HHEAD * HBASE
assert model_dict['HSIZE'] == model_dict['HHEAD'] * model_dict['HBASE']

cfg = Config()

# Create input tensor: (batch_size=2, seq_len=3, hidden_size=8)
x = torch.randn(2, 3, model_dict['HSIZE'])

# Create mask tensor for attention: (batch_size=2, num_heads=2, seq_len=3, seq_len=3)
# Use zeros for a valid mask (no masking), or negative inf where we want to mask
x_mask = torch.zeros(2, model_dict['HHEAD'], 3, 3)

ARGS = {
    'cfg': cfg,
    'model_dict': model_dict
}
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 Backbone — 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 Backbone 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
FeedForward Ran
ran on a synthesized input · constructed an object; no result was computed
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("1e4945df552f6045")
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, FeedForward: 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
import math

torch.manual_seed(0)

# Define a minimal config dict
cfg = {
    'model_name': 'test_model',
    'vocab_size': 100,
}

# Define model_dict with required keys used in FeedForward
model_dict = {
    'HSIZE': 4,      # Hidden size (input/output dimension)
    'HFF': 8,        # Hidden feed-forward dimension
}

# Construct ARGS mapping all parameters
ARGS = {
    'cfg': cfg,
    'model_dict': model_dict,
}
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 FeedForward — 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 FeedForward 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
LayerNorm Ran
ran on a synthesized input · constructed an object; no result was computed
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("fcab94bf5f3f29d5")
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, LayerNorm: 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)

# Create a simple input tensor for testing
# Shape: (batch_size=2, sequence_length=3, hidden_size=4)
x = torch.randn(2, 3, 4)

# Define ARGS dict with all parameters for LayerNorm
ARGS = {
    "hidden_size": 4,
    "eps": 1e-12
}
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 LayerNorm — 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 LayerNorm 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
Pooler Ran
ran on a synthesized input · constructed an object; no result was computed
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("14d39fe285bd537f")
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, Pooler: 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)

class Config:
    def __init__(self, hsize):
        self.HSIZE = hsize

cfg = Config(hsize=4)

ARGS = {
    'cfg': cfg
}
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 Pooler — 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 Pooler 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
TextEmbeddings Ran
ran on a synthesized input · constructed an object; no result was computed
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("1e2c2b0e7e10968a")
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, TextEmbeddings: 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)

class Config:
    def __init__(self):
        self.HSIZE = 8
        self.POS_EMB_IN_SIZE = 16
        self.TYPE_EMB_IN_SIZE = 2
        self.DROPOUT_R = 0.1

cfg = Config()
vocab_size = 32

input_ids = torch.randint(0, vocab_size, (2, 4), dtype=torch.long)

ARGS = {
    "cfg": cfg,
    "vocab_size": vocab_size
}
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 TextEmbeddings — 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 TextEmbeddings 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
VisualEmbeddings Ran
ran on a synthesized input · constructed an object; no result was computed
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("b0d904c1669cb4b0")
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, VisualEmbeddings: 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)

class Config:
    def __init__(self):
        self.IMGFEAT_SIZE = 5
        self.BBOXFEAT_SIZE = 4
        self.HSIZE = 6
        self.USE_BBOXFEAT = True
        self.DROPOUT_R = 0.1

cfg = Config()

ARGS = {
    'cfg': cfg
}
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 VisualEmbeddings — 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 VisualEmbeddings 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
gelu Ran
ran, honoured the contract we drafted
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("8d619e54294fb0b7")
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 torch
import math

torch.manual_seed(0)

# The gelu function takes a single parameter 'x' which is used with:
# - torch.erf() - requires a tensor
# - math.sqrt() - standard math function
# - arithmetic operations (*, +, /)
# 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
get_list_text Ran
ran on a synthesized input
ThanThoai/Visual-Question-Answering_Vietnamese/merge_raw_data.py
code served (permissive licence) · get_code("b9f7aa42c0afe2be")
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 temporary directories with test files
temp_dir = tempfile.mkdtemp()
path_en = os.path.join(temp_dir, 'en')
path_vi = os.path.join(temp_dir, 'vi')
path_idx = os.path.join(temp_dir, 'idx')

os.makedirs(path_en, exist_ok=True)
os.makedirs(path_vi, exist_ok=True)
os.makedirs(path_idx, exist_ok=True)

# Create test files for 'train' type
# File: dataset_train.txt (English)
with open(os.path.join(path_en, 'dataset_train.txt'), 'w') as f:
    f.write("hello world\n")
    f.write("how are you\n")

# File: dataset_idx_train.txt (Index)
with open(os.path.join(path_idx, 'dataset_idx_train.txt'), 'w') as f:
    f.write("0\n")
    f.write("1\n")

# File: dataset.en.vi.txt (Vietnamese)
with open(os.path.join(path_vi, 'dataset.en.vi.txt'), 'w') as f:
    f.write("xin chào thế giới\n")
    f.write("bạn khỏe không\n")

ARGS = {
    'path_en': path_en,
    'path_vi': path_vi,
    'path_idx': path_idx,
    'type_': 'train'
}
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
make_data_sampler Ran
ran on a synthesized input
microsoft/Oscar/oscar/run_captioning.py
code served (permissive licence) · get_code("e17f1049b8148f71")
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 RandomSampler. 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.utils.data import Dataset

torch.manual_seed(0)

class DummyDataset(Dataset):
    def __init__(self, size=10):
        self.size = size
    
    def __len__(self):
        return self.size
    
    def __getitem__(self, idx):
        return torch.randn(4), torch.tensor(idx % 2)

dataset = DummyDataset(size=10)

ARGS = {
    "dataset": dataset,
    "shuffle": True,
    "distributed": False
}
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 RandomSampler — a RandomSampler output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a RandomSampler 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
make_mask Ran
ran; our contract draft was wrong, not the code
ThanThoai/Visual-Question-Answering_Vietnamese/model.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from MRUIL/WeakMCN, is pointer only (licence: NONE) · get_code("e766a0a3063c15d0")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from MRUIL/WeakMCN, 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.

optimal_transport_dist Ran
ran; our contract draft was wrong, not the code
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("06b045032beeb885")
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 1, dtype family float, shape (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create small tensors for testing
batch_size = 2
txt_seq_len = 3
img_seq_len = 4
embedding_dim = 5

# Text embeddings: [B, M, D]
txt_emb = torch.randn(batch_size, txt_seq_len, embedding_dim)

# Image embeddings: [B, N, D]
img_emb = torch.randn(batch_size, img_seq_len, embedding_dim)

# Text padding mask: [B, M] - False means valid, True means padded
txt_pad = torch.tensor([
    [False, False, True],
    [False, False, False]
])

# Image padding mask: [B, N] - False means valid, True means padded
img_pad = torch.tensor([
    [False, False, False, True],
    [False, False, True, True]
])

ARGS = {
    "txt_emb": txt_emb,
    "img_emb": img_emb,
    "txt_pad": txt_pad,
    "img_pad": img_pad,
    "beta": 0.5,
    "iteration": 50,
    "k": 1
}
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 1, dtype family float, shape (2,) — shape (2,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
read_json Ran
ran on a synthesized input
ThanThoai/Visual-Question-Answering_Vietnamese/merge_raw_data.py
code served (permissive licence) · get_code("ecf115f3542eefd4")
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 json
import tempfile
import os

# Create a temporary JSON file with sample data
temp_dir = tempfile.mkdtemp()
json_path = os.path.join(temp_dir, "test.json")

sample_data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "key3": {"nested": "object"}
}

with open(json_path, 'w') as f:
    json.dump(sample_data, f)

ARGS = {
    "path_json": json_path
}
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 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
trace Ran
ran; our contract draft was wrong, not the code
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("412cdf6ed5178c2c")
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 1, dtype family float, shape (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a small batched square matrix tensor
# Shape: (batch=2, m=4, n=4)
x = torch.randn(2, 4, 4)

ARGS = {
    "x": x
}
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 1, dtype family float, shape (2,) — shape (2,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
ImgfeatHead Not yet run
unverified (harvested, no recorded run)
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("01007aebf0b1b84a")
How it ran

no invocation recorded.

MHAtt Not yet run
unverified (harvested, no recorded run)
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("b97a0ae0d5f606f5")
How it ran

no invocation recorded.

MmITMHead Not yet run
unverified (harvested, no recorded run)
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("8bc6b6ae220fc251")
How it ran

no invocation recorded.

MmQAHead Not yet run
unverified (harvested, no recorded run)
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("b90f9bcd35cb01ad")
How it ran

no invocation recorded.

Net Not yet run
unverified (harvested, no recorded run)
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("4fde87aac2867c6f")
How it ran

no invocation recorded.

SelfAtt Not yet run
unverified (harvested, no recorded run)
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("00f7c9f9cc6a5223")
How it ran

no invocation recorded.

TextMLMHead Not yet run
unverified (harvested, no recorded run)
milvlg/rosita/rosita/modeling/pretrain_tasks/rosita.py
code served (permissive licence) · get_code("5b65287416e63add")
How it ran

no invocation recorded.

get_data_en Not yet run
unverified (harvested, no recorded run)
ThanThoai/Visual-Question-Answering_Vietnamese/merge_raw_data.py
code served (permissive licence) · get_code("5d4575f5eacfe1f9")
How it ran

no invocation recorded.

prepare_data Not yet run
unverified (harvested, no recorded run)
ThanThoai/Visual-Question-Answering_Vietnamese/extract_image.py
code served (permissive licence) · get_code("e6bf616333cbcc3b")
How it ran

no invocation recorded.

train_model Not yet run
unverified (harvested, no recorded run)
ThanThoai/Visual-Question-Answering_Vietnamese/extract_image.py
code served (permissive licence) · get_code("bdc984dffbfed9a3")
How it ran

no invocation recorded.

Also present in other papers' code

Of 23 samples lifted from this paper's linked repositories, 23 carry a function identity and 0 do not. Of the 23 distinct functions with an identity, 4 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 2 the same month, 4 later) and 19 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
gelu
brick of this paper's assembly: helper
198 other papers
38 dated earlier, 8 the same month, 150 later, 2 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
make_mask 7 other papers
7 later
2006.15955 (2020-06)
A Transformer-based joint-encoding for Emotion Recognition and Sentiment Analysis
2305.14836 (2023-05)
NuScenes-QA: A Multi-modal Visual Question Answering Benchmark for Autonomous Driving Scenario
2307.05182 (2023-07)
CAT-ViL: Co-Attention Gated Vision-Language Embedding for Visual Question Localized-Answering in Robotic Surgery
and 4 more not named here
LayerNorm
brick of this paper's assembly: helper
5 other papers
1 dated earlier, 4 later · 3 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
2005.05806 (2020-05)
Document Modeling with Graph Attention Networks for Multi-grained Machine Reading Comprehension
2401.06633 (2024-01)
Ada-Retrieval: An Adaptive Multi-Round Retrieval Paradigm for Sequential Recommendations
and 2 more not named here
make_data_sampler 4 other papers
1 the same month, 3 later
2004.01946 (2020-04)
Weakly-Supervised Mesh-Convolutional Hand Reconstruction in the Wild
2101.00529 (2021-01)
VinVL: Revisiting Visual Representations in Vision-Language Models
2211.09371 (2022-11)
CapEnrich: Enriching Caption Semantics for Web Images via Cross-modal Pre-trained Knowledge
and 1 more not named here

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-21.

Repositories linked to this paper

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

Abstract

Large-scale pre-training methods of learning cross-modal representations on image-text pairs are becoming popular for vision-language tasks. While existing methods simply concatenate image region features and text features as input to the model to be pre-trained and use selfattention to learn image-text semantic alignments in a brute force manner, in this paper, we propose a new learning method Oscar 1 , which uses object tags detected in images as anchor points to significantly ease the learning of alignments. Our method is motivated by the observation that the salient objects in an image can be accurately detected, and are often mentioned in the paired text. We pre-train an Oscar model on the public corpus of 6.5 million text-image pairs, and fine-tune it on downstream tasks, creating new state-of-the-arts on six well-established vision-language understanding and generation tasks. 2

For agents

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

get_harvested_code_for_paper("2004.06165")
get_code_for_paper("2004.06165")
have("2004.06165")

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

curl https://syntology.ai/api/ran/2004.06165.json

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

[![Syntology run record](https://syntology.ai/api/ran/2004.06165.svg)](https://syntology.ai/paper/2004.06165)

Connect an agent — have() is free.