SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2006.06195 · NeurIPS · 2020

Large-Scale Adversarial Training for Vision-and-Language Representation Learning

A Research, Chen Zhu, Yu Cheng, Zhe Gan, Yen-Chun Chen, Linjie Li, Jingjing Liu, Microsoft Dynamics

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 (5 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 3 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/2006.06195.json · badge: /api/ran/2006.06195.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
zhegan27/VILLA canonical (classifier label)
classifier basis: census
6 ran · 3 no recorded run
zhegan27/LXMERT-AdvTrain marked official in Papers with Code
Papers with Code archive
6 ran · 5 no recorded run
FunctionStatusWhere it lives
swish Ran
ran; our contract draft was wrong, not the code
zhegan27/VILLA/model/layer.py
pointer only (licence: MIT) · get_code("0f786c407fb1ee4c")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, zhegan27/VILLA, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. 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.

BertEmbeddings Ran
ran on a synthesized input
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("7cd6bab68e15976d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 3, 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
from torch import nn

BertLayerNorm = torch.nn.LayerNorm

class BertEmbeddings(nn.Module):
    """Construct the embeddings from word, position and token_type embeddings.
    """

    def __init__(self, config):
        super(BertEmbeddings, self).__init__()
        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=0)
        self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size, padding_idx=0)
        self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size, padding_idx=0)
        self.LayerNorm = BertLayerNorm(config.hidden_size, eps=1e-12)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)

    def forward(self, input_ids, token_type_ids=None, adv_training=False, adv_modality=None, adv_delta=None):
        seq_length = input_ids.size(1)
        position_ids = torch.arange(seq_length, dtype=torch.long, device=input_ids.device)
        position_ids = position_ids.unsqueeze(0).expand_as(input_ids)
        if token_type_ids is None:
            token_type_ids = torch.zeros_like(input_ids)
        words_embeddings = self.word_embeddings(input_ids)
        if adv_training == True and 'text' in adv_modality:
            words_embeddings = words_embeddings + adv_delta
        position_embeddings = self.position_embeddings(position_ids)
        token_type_embeddings = self.token_type_embeddings(token_type_ids)
        embeddings = words_embeddings + position_embeddings + token_type_embeddings
        embeddings = self.LayerNorm(embeddings)
        embeddings = self.dropout(embeddings)
        return embeddings


class Config:
    def __init__(self):
        self.vocab_size = 100
        self.hidden_size = 8
        self.max_position_embeddings = 16
        self.type_vocab_size = 3
        self.hidden_dropout_prob = 0.1


INIT_ARGS = {
    'config': Config()
}

ARGS = {
    'input_ids': torch.randint(0, 100, (2, 5), dtype=torch.long),
    'token_type_ids': torch.randint(0, 3, (2, 5), dtype=torch.long),
    'adv_training': False,
    'adv_modality': None,
    'adv_delta': None
}

if __name__ == '__main__':
    model = BertEmbeddings(**INIT_ARGS)
    output = model(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Output dtype: {output.dtype}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 3, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
BertPredictionHeadTransform Ran
ran on a synthesized input · constructed an object; no result was computed
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("2ba3b1dfb07ba4b1")
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, BertPredictionHeadTransform: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math
import sys
import torch
from torch import nn

torch.manual_seed(0)

class Config:
    def __init__(self, hidden_size=4, hidden_act='gelu'):
        self.hidden_size = hidden_size
        self.hidden_act = hidden_act

config = Config(hidden_size=4, hidden_act='gelu')

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 BertPredictionHeadTransform — 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 BertPredictionHeadTransform 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
BertVisualAnswerHead Ran
ran on a synthesized input · constructed an object; no result was computed
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("2c8c022f71d9ae6e")
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, BertVisualAnswerHead: 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)

class MockConfig:
    def __init__(self, hidden_size):
        self.hidden_size = hidden_size

config = MockConfig(hidden_size=4)
num_answers = 3

hidden_states = torch.randn(2, 4)

ARGS = {
    "config": config,
    "num_answers": num_answers
}
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 BertVisualAnswerHead — 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 BertVisualAnswerHead 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
BertVisualObjHead Ran
ran on a synthesized input · constructed an object; no result was computed
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("0e7be2781b71b508")
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, BertVisualObjHead: 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
import sys

torch.manual_seed(0)

def gelu(x):
    return x * 0.5 * (1.0 + torch.erf(x / math.sqrt(2.0)))

def swish(x):
    return x * torch.sigmoid(x)

ACT2FN = {'gelu': gelu, 'relu': torch.nn.functional.relu, 'swish': swish}

class VisualConfig(object):
    VISUAL_LOSSES = ['obj', 'attr', 'feat']

    def __init__(self, l_layers=12, x_layers=5, r_layers=0):
        self.l_layers = l_layers
        self.x_layers = x_layers
        self.r_layers = r_layers
        self.visual_feat_dim = 2048
        self.visual_pos_dim = 4
        self.obj_id_num = 1600
        self.attr_id_num = 400
        self.visual_losses = self.VISUAL_LOSSES
        self.visual_loss_config = {
            'obj': (self.obj_id_num, 'ce', (-1,), 1 / 0.15),
            'attr': (self.attr_id_num, 'ce', (-1,), 1 / 0.15),
            'feat': (2048, 'l2', (-1, 2048), 1 / 0.15)
        }

    def set_visual_dims(self, feat_dim, pos_dim):
        self.visual_feat_dim = feat_dim
        self.visual_pos_dim = pos_dim

VISUAL_CONFIG = VisualConfig()

class BertConfig(object):
    def __init__(self, hidden_size=768, hidden_act='gelu'):
        self.hidden_size = hidden_size
        self.hidden_act = hidden_act

config = BertConfig(hidden_size=256, hidden_act='gelu')
visual_losses = 'obj,attr'

ARGS = {
    'config': config,
    'visual_losses': visual_losses
}
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 BertVisualObjHead — 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 BertVisualObjHead 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 on a synthesized input · constructed an object; no result was computed
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("ce655ad320784ce8")
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, GeLU: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math
import torch
from torch import nn

torch.manual_seed(0)

class GeLU(nn.Module):
    """Implementation of the gelu activation function.
        For information: OpenAI GPT's gelu is slightly different (and gives slightly different results):
        0.5 * x * (1 + torch.tanh(math.sqrt(2 / math.pi) * (x + 0.044715 * torch.pow(x, 3))))
        Also see https://arxiv.org/abs/1606.08415
    """

    def __init__(self):
        super().__init__()

    def forward(self, x):
        return x * 0.5 * (1.0 + torch.erf(x / math.sqrt(2.0)))

ARGS = {}
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 GeLU — 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 GeLU 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
VisualFeatEncoder Ran
ran on a synthesized input · constructed an object; no result was computed
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("1bea2e32e2dc619d")
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, VisualFeatEncoder: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch import nn

torch.manual_seed(0)

# Create a minimal config class that has the required attributes
class MinimalConfig:
    def __init__(self):
        self.hidden_size = 768
        self.hidden_dropout_prob = 0.1

config = MinimalConfig()

# Create tiny tensors for the forward pass
# feats: [batch_size, num_objects, visual_feat_dim]
# boxes: [batch_size, num_objects, visual_pos_dim]
batch_size = 2
num_objects = 3
feats = torch.randn(batch_size, num_objects, 2048)  # visual_feat_dim = 2048
boxes = torch.randn(batch_size, num_objects, 4)     # visual_pos_dim = 4

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 VisualFeatEncoder — 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 VisualFeatEncoder 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
compute_score_with_logits Ran
ran on a synthesized input
zhegan27/VILLA/inf_vqa.py
pointer only (licence: MIT) · get_code("2731fd9de6d31bc0")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, zhegan27/VILLA, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. 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.

cost_matrix_cosine Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zhegan27/VILLA/model/ot.py
pointer only (licence: MIT); the stored copy get_code serves, from limanling/clip-event, is pointer only (licence: NONE) · get_code("9c9d2823696faa43")
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 limanling/clip-event, 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
zhegan27/VILLA/model/layer.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: MIT, as GitHub detects it on this paper's repository, zhegan27/VILLA, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. 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.

multi_head_attention_forward Ran
ran; our contract draft was wrong, not the code
zhegan27/VILLA/model/attention.py
code served (permissive licence) · get_code("fed42c63be2ceb6c")
How it ran

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

import torch
from torch.nn.functional import linear, softmax, dropout

torch.manual_seed(0)

# Parameters
batch_size = 2
seq_len_query = 3
seq_len_key = 5
embed_dim = 8
num_heads = 2
head_dim = embed_dim // num_heads

# Construct tensors
query = torch.randn(seq_len_query, batch_size, embed_dim)
key = torch.randn(seq_len_key, batch_size, embed_dim)
value = torch.randn(seq_len_key, batch_size, embed_dim)

# Input projection weights and biases
# Combined weight for q, k, v projections
in_proj_weight = torch.randn(3 * embed_dim, embed_dim)
in_proj_bias = torch.randn(3 * embed_dim)

# Output projection
out_proj_weight = torch.randn(embed_dim, embed_dim)
out_proj_bias = torch.randn(embed_dim)

# Optional parameters
bias_k = None
bias_v = None
add_zero_attn = False
dropout_p = 0.1
key_padding_mask = None
attn_mask = None
need_weights = True
use_separate_proj_weight = False
q_proj_weight = None
k_proj_weight = None
v_proj_weight = None
static_k = None
static_v = None
training = True

ARGS = {
    "query": query,
    "key": key,
    "value": value,
    "embed_dim_to_check": embed_dim,
    "num_heads": num_heads,
    "in_proj_weight": in_proj_weight,
    "in_proj_bias": in_proj_bias,
    "bias_k": bias_k,
    "bias_v": bias_v,
    "add_zero_attn": add_zero_attn,
    "dropout_p": dropout_p,
    "out_proj_weight": out_proj_weight,
    "out_proj_bias": out_proj_bias,
    "training": training,
    "key_padding_mask": key_padding_mask,
    "need_weights": need_weights,
    "attn_mask": attn_mask,
    "use_separate_proj_weight": use_separate_proj_weight,
    "q_proj_weight": q_proj_weight,
    "k_proj_weight": k_proj_weight,
    "v_proj_weight": v_proj_weight,
    "static_k": static_k,
    "static_v": static_v,
}
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 tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
save_for_submission Ran
ran on a synthesized input
zhegan27/VILLA/inf_vcr.py
pointer only (licence: MIT) · get_code("70361102cdeaaf85")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, zhegan27/VILLA, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. 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.

BertLMPredictionHead Not yet run
unverified (harvested, no recorded run)
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("77ff902ef6858211")
How it ran

no invocation recorded.

BertPreTrainingHeads Not yet run
unverified (harvested, no recorded run)
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("781e8b75f1ca0f7c")
How it ran

no invocation recorded.

LXRTEncoder Not yet run
unverified (harvested, no recorded run)
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("2c0fd477bfbddac7")
How it ran

no invocation recorded.

LXRTModel Not yet run
unverified (harvested, no recorded run)
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("95e61302335909fb")
How it ran

no invocation recorded.

LXRTPretraining Not yet run
unverified (harvested, no recorded run)
zhegan27/LXMERT-AdvTrain/src/lxrt/modeling.py
code served (permissive licence) · get_code("6f76d36162ffef27")
How it ran

no invocation recorded.

evaluate Not yet run
unverified (harvested, no recorded run)
zhegan27/VILLA/inf_nlvr2.py
pointer only (licence: MIT) · get_code("28e194bd5c984ae7")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, zhegan27/VILLA, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. 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. No invocation recorded.

ipot Not yet run
unverified (harvested, no recorded run)
zhegan27/VILLA/model/ot.py
pointer only (licence: MIT); the stored copy get_code serves, from limanling/clip-event, is pointer only (licence: NONE) · get_code("cbf43203110456f0")
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 limanling/clip-event, 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. No invocation recorded.

trace Not yet run
unverified (harvested, no recorded run)
zhegan27/VILLA/model/ot.py
pointer only (licence: MIT) · get_code("d3f4c62a97861b9d")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, zhegan27/VILLA, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. 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. 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, 14 are also present in other papers whose code we hold (11 with at least one paper dated earlier, 2 the same month, 9 later) and 6 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
brick of this paper's assembly: helper
206 other papers
39 dated earlier, 2 the same month, 164 later, 1 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
brick of this paper's assembly: helper
198 other papers
53 dated earlier, 1 the same month, 142 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
BertLMPredictionHead
brick of this paper's assembly: model
3 other papers
2 dated earlier, 1 later · 1 only as an assembly brick
1908.02265 (2019-08)
ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks
1909.10351 (2019-09)
TinyBERT: Distilling BERT for Natural Language Understanding
2107.14596 (2021-07)
Multi-stage Pre-training over Simplified Multimodal Pre-training Models
BertPredictionHeadTransform
brick of this paper's assembly: model
3 other papers
2 dated earlier, 1 later · 1 only as an assembly brick
1908.02265 (2019-08)
ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks
1909.10351 (2019-09)
TinyBERT: Distilling BERT for Natural Language Understanding
2107.14596 (2021-07)
Multi-stage Pre-training over Simplified Multimodal Pre-training Models
GeLU
brick of this paper's assembly: helper
2 other papers
2 later · 2 only as an assembly brick
2107.14596 (2021-07)
Multi-stage Pre-training over Simplified Multimodal Pre-training Models
2304.08682 (2023-04)
Learning Situation Hyper-Graphs for Video Question Answering
cost_matrix_cosine 2 other papers
1 dated earlier, 1 later · 1 only as an assembly brick
2004.06165 (2020-04)
Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks
2201.05078 (2022-01)
CLIP-Event: Connecting Text and Images with Event Structures
ipot 2 other papers
1 dated earlier, 1 later · 1 only as an assembly brick
2004.06165 (2020-04)
Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks
2201.05078 (2022-01)
CLIP-Event: Connecting Text and Images with Event Structures
BertEmbeddings
brick of this paper's assembly: model
1 other paper
1 dated earlier · 1 only as an assembly brick
1908.07490 (2019-08)
LXMERT: Learning Cross-Modality Encoder Representations from Transformers
BertPreTrainingHeads
brick of this paper's assembly: model
1 other paper
1 dated earlier
1909.10351 (2019-09)
TinyBERT: Distilling BERT for Natural Language Understanding
BertVisualAnswerHead
brick of this paper's assembly: model
1 other paper
1 later · 1 only as an assembly brick
2107.14596 (2021-07)
Multi-stage Pre-training over Simplified Multimodal Pre-training Models
LXRTEncoder
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
1908.07490 (2019-08)
LXMERT: Learning Cross-Modality Encoder Representations from Transformers
LXRTModel
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
1908.07490 (2019-08)
LXMERT: Learning Cross-Modality Encoder Representations from Transformers
VisualFeatEncoder
brick of this paper's assembly: model
1 other paper
1 dated earlier · 1 only as an assembly brick
1908.07490 (2019-08)
LXMERT: Learning Cross-Modality Encoder Representations from Transformers
multi_head_attention_forward 1 other paper
1 later
2105.13868 (2021-05)
Learning Relation Alignment for Calibrated Cross-modal Retrieval

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

We present VILLA, the first known effort on large-scale adversarial training for vision-and-language (V+L) representation learning. VILLA consists of two training stages: (i) task-agnostic adversarial pre-training; followed by (ii) task-specific adversarial finetuning. Instead of adding adversarial perturbations on image pixels and textual tokens, we propose to perform adversarial training in the embedding space of each modality. To enable large-scale training, we adopt the "free" adversarial training strategy, and combine it with KL-divergence-based regularization to promote higher invariance in the embedding space. We apply VILLA to current best-performing V+L models, and achieve new state of the art on a wide range of tasks, including Visual Question Answering, Visual Commonsense Reasoning, Image-Text Retrieval, Referring Expression Comprehension, Visual Entailment, and NLVR 2 . 1

For agents

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

get_harvested_code_for_paper("2006.06195")
get_code_for_paper("2006.06195")
have("2006.06195")

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

curl https://syntology.ai/api/ran/2006.06195.json

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

[![Syntology run record](https://syntology.ai/api/ran/2006.06195.svg)](https://syntology.ai/paper/2006.06195)

Connect an agent — have() is free.