SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2207.05342 · ECCV · 2022

Video Graph Transformer for Video Question Answering

Pan Zhou, Shuicheng Yan, Tat-Seng Chua, Junbin Xiao

arXiv · PDF · Open in the Atlas

Code that ran

9 of 14 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; 0 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 1 failure of our instrument; 5 no recorded run), 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/2207.05342.json · badge: /api/ran/2207.05342.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
sail-sg/VGT community
Papers with Code archive
9 ran · 5 no recorded run
FunctionStatusWhere it lives
CMAtten Ran
ran on a synthesized input
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("f3ff51937dd98c21")
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 tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable

# INIT_ARGS for CMAtten.__init__
INIT_ARGS = {}

# ARGS for CMAtten.forward
# s1: [B, t1, D]
# l1: [B] - lengths for s1
# s2: [B, t2, D]
# l2: [B] - lengths for s2

B = 2
t1 = 5
t2 = 7
D = 4

s1 = torch.randn(B, t1, D)
l1 = torch.tensor([3, 4])  # actual lengths for each batch element
s2 = torch.randn(B, t2, D)
l2 = torch.tensor([5, 6])  # actual lengths for each batch element

ARGS = {
    's1': s1,
    'l1': l1,
    's2': s2,
    'l2': l2
}
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 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
Embeddings Ran
ran on a synthesized input · constructed an object; no result was computed
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("5a6a727d5b0f9c6b")
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, Embeddings: 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 numpy as np

torch.manual_seed(0)

ARGS = {
    "d_model": 4,
    "language_len": 3,
    "vision_len": 5,
    "dropout": 0.1,
    "sinusoidal_pos_embds": True,
    "d_pos": 128,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Embeddings — 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 Embeddings 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
EncoderVid Ran
ran on a synthesized input · constructed an object; no result was computed
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("66dcb62d12860979")
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, EncoderVid: 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)

# Parameters for EncoderVid
feat_dim = 4
bbox_dim = 3
feat_hidden = 8
pos_hidden = 5
input_dropout_p = 0.3

# Create input tensor with shape (bsize, numc, numf, numr, fdim)
# where fdim = feat_dim + bbox_dim (features + bounding boxes)
bsize = 2
numc = 2
numf = 3
numr = 4
video_o = torch.randn(bsize, numc, numf, numr, feat_dim + bbox_dim)

ARGS = {
    'feat_dim': feat_dim,
    'bbox_dim': bbox_dim,
    'feat_hidden': feat_hidden,
    'pos_hidden': pos_hidden,
    'input_dropout_p': input_dropout_p,
}
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 EncoderVid — 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 EncoderVid 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
FFN Ran
ran on a synthesized input · constructed an object; no result was computed
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("fad49f16c43d52df")
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, FFN: 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
from transformers.activations import gelu

torch.manual_seed(0)

class Config:
    def __init__(self):
        self.attention_probs_dropout_prob = 0.1
        self.hidden_size = 4
        self.intermediate_size = 8
        self.hidden_act = 'gelu'

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, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type FFN — 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 FFN 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
Graph Ran
ran on a synthesized input · constructed an object; no result was computed
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("c7325ef7876ef60e")
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, Graph: 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)

ARGS = {
    "dim_in": 4,
    "dim_hidden": 8,
    "dim_out": 6,
    "num_layers": 2,
    "dropout": 0.5,
}
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 Graph — 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 Graph 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
GraphConvolution Ran
ran on a synthesized input · constructed an object; no result was computed
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("889a03d6a437ba1c")
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, GraphConvolution: 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
from torch.nn.parameter import Parameter

torch.manual_seed(0)

class GraphConvolution(nn.Module):
    """
    Simple GCN layer, similar to https://arxiv.org/abs/1609.02907
    """

    def __init__(self, in_features, out_features, bias=True, skip=True):
        super(GraphConvolution, self).__init__()
        self.skip = skip
        self.in_features = in_features
        self.out_features = out_features
        self.weight = Parameter(torch.Tensor(in_features, out_features))
        if bias:
            self.bias = Parameter(torch.Tensor(out_features))
        else:
            self.register_parameter('bias', None)
        self.reset_parameters()

    def reset_parameters(self):
        stdv = 1.0 / math.sqrt(self.weight.size(1))
        self.weight.data.uniform_(-stdv, stdv)
        if self.bias is not None:
            self.bias.data.uniform_(-stdv, stdv)

    def forward(self, input, adj):
        support = torch.bmm(input, self.weight.unsqueeze(0).expand(input.shape[0], -1, -1))
        output = torch.bmm(adj, support)
        if self.bias is not None:
            output += self.bias.unsqueeze(0).expand(input.shape[0], -1, -1)
        if self.skip:
            output += support
        return output

    def __repr__(self):
        return self.__class__.__name__ + ' (' + str(self.in_features) + ' -> ' + str(self.out_features) + ')'


# Create test inputs
batch_size = 2
in_features = 4
out_features = 3
num_nodes = 5

# input tensor: (batch_size, num_nodes, in_features)
input_tensor = torch.randn(batch_size, num_nodes, in_features)

# adj tensor: (batch_size, num_nodes, num_nodes)
adj_tensor = torch.randn(batch_size, num_nodes, num_nodes)

ARGS = {
    'in_features': in_features,
    'out_features': out_features,
    'bias': True,
    'skip': True,
}
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 GraphConvolution — 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 GraphConvolution 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
MultiHeadSelfAttention Ran
ran on a synthesized input
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("b091b2a22b3049e3")
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 tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

class Config:
    def __init__(self):
        self.num_attention_heads = 2
        self.hidden_size = 8
        self.attention_probs_dropout_prob = 0.1

config = Config()

INIT_ARGS = {
    'config': config
}

batch_size = 2
seq_length = 5
hidden_size = config.hidden_size

query = torch.randn(batch_size, seq_length, hidden_size)
key = torch.randn(batch_size, seq_length, hidden_size)
value = torch.randn(batch_size, seq_length, hidden_size)
mask = torch.ones(batch_size, seq_length)

ARGS = {
    'query': query,
    'key': key,
    'value': value,
    'mask': mask,
    'head_mask': None,
    'output_attentions': False
}
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 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
TransformerBlock Ran
ran on a synthesized input
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("b294b4902336fb78")
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 tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
from transformers.activations import gelu
import math

class FFN(nn.Module):

    def __init__(self, config):
        super().__init__()
        dropout, dim, hidden_dim = (config.attention_probs_dropout_prob, config.hidden_size, config.intermediate_size)
        activation = config.hidden_act
        self.dropout = nn.Dropout(p=dropout)
        self.lin1 = nn.Linear(in_features=dim, out_features=hidden_dim)
        self.lin2 = nn.Linear(in_features=hidden_dim, out_features=dim)
        assert activation in ['relu', 'gelu'], "activation ({}) must be in ['relu', 'gelu']".format(activation)
        self.activation = gelu if activation == 'gelu' else nn.ReLU()

    def forward(self, input):
        x = self.lin1(input)
        x = self.activation(x)
        x = self.lin2(x)
        x = self.dropout(x)
        return x

class MultiHeadSelfAttention(nn.Module):

    def __init__(self, config):
        super().__init__()
        self.n_heads = config.num_attention_heads
        self.dim = config.hidden_size
        dp_rate = config.attention_probs_dropout_prob
        self.dropout = nn.Dropout(p=dp_rate)
        assert self.dim % self.n_heads == 0
        self.q_lin = nn.Linear(in_features=self.dim, out_features=self.dim)
        self.k_lin = nn.Linear(in_features=self.dim, out_features=self.dim)
        self.v_lin = nn.Linear(in_features=self.dim, out_features=self.dim)
        self.out_lin = nn.Linear(in_features=self.dim, out_features=self.dim)
        self.pruned_heads = set()

    def forward(self, query, key, value, mask, head_mask=None, output_attentions=False):
        bs, q_length, dim = query.size()
        k_length = key.size(1)
        dim_per_head = self.dim // self.n_heads
        mask_reshp = (bs, 1, 1, k_length)

        def shape(x):
            return x.view(bs, -1, self.n_heads, dim_per_head).transpose(1, 2)

        def unshape(x):
            return x.transpose(1, 2).contiguous().view(bs, -1, self.n_heads * dim_per_head)
        q = shape(self.q_lin(query))
        k = shape(self.k_lin(key))
        v = shape(self.v_lin(value))
        q = q / math.sqrt(dim_per_head)
        scores = torch.matmul(q, k.transpose(2, 3))
        mask = (mask == 0).view(mask_reshp).expand_as(scores)
        scores.masked_fill_(mask, -float('inf'))
        weights = nn.Softmax(dim=-1)(scores)
        weights = self.dropout(weights)
        if head_mask is not None:
            weights = weights * head_mask
        context = torch.matmul(weights, v)
        context = unshape(context)
        context = self.out_lin(context)
        if output_attentions:
            return (context, weights)
        else:
            return (context,)

class TransformerBlock(nn.Module):

    def __init__(self, config):
        super().__init__()
        dim = config.hidden_size
        assert dim % config.num_attention_heads == 0
        self.attention = MultiHeadSelfAttention(config)
        self.sa_layer_norm = nn.LayerNorm(normalized_shape=dim, eps=1e-12)
        self.ffn = FFN(config)
        self.output_layer_norm = nn.LayerNorm(normalized_shape=dim, eps=1e-12)

    def forward(self, x, attn_mask=None, head_mask=None, output_attentions=False):
        sa_output = self.attention(query=x, key=x, value=x, mask=attn_mask, head_mask=head_mask, output_attentions=output_attentions)
        if output_attentions:
            sa_output, sa_weights = sa_output
        else:
            assert type(sa_output) == tuple
            sa_output = sa_output[0]
        sa_output = self.sa_layer_norm(sa_output + x)
        ffn_output = self.ffn(sa_output)
        ffn_output = self.output_layer_norm(ffn_output + sa_output)
        output = (ffn_output,)
        if output_attentions:
            output = (sa_weights,) + output
        return output


class Config:
    def __init__(self):
        self.hidden_size = 4
        self.num_attention_heads = 2
        self.intermediate_size = 8
        self.attention_probs_dropout_prob = 0.1
        self.hidden_act = 'gelu'


INIT_ARGS = {
    'config': Config()
}

ARGS = {
    'x': torch.randn(2, 3, 4),
    'attn_mask': torch.ones(2, 3),
    'head_mask': None,
    'output_attentions': False
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
torch, transformers — 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
get_mask Ran
ran; our contract draft was wrong, not the code
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("4b0473dade9942bc")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family int, 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

torch.manual_seed(0)

# lengths: a 1D tensor of sequence lengths for each item in the batch
lengths = torch.tensor([3, 2], dtype=torch.long)

# max_length: scalar integer representing the maximum sequence length
max_length = 4

ARGS = {
    "lengths": lengths,
    "max_length": max_length,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family int, 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
AModel Not yet run
unverified (harvested, no recorded run)
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("2f0df7b215389abb")
How it ran

no invocation recorded.

Bert Not yet run
unverified (harvested, no recorded run)
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("f64528547183d480")
How it ran

no invocation recorded.

Transformer Not yet run
unverified (harvested, no recorded run)
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("9d80cb2ca7b058a1")
How it ran

no invocation recorded.

VGT Not yet run
unverified (harvested, no recorded run)
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("cdedc918d8a5dc8b")
How it ran

no invocation recorded.

create_sinusoidal_embeddings Not yet run
unverified (harvested, no recorded run)
sail-sg/VGT/model/vqa_model.py
code served (permissive licence) · get_code("741ba8df50c491fc")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 13 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
get_mask
brick of this paper's assembly: helper
2 other papers
2 later
2307.12058 (2023-07)
Discovering Spatio-Temporal Rationales for Video Question Answering
2308.09363 (2023-08)
Open-Vocabulary Video Question Answering: A New Benchmark for Evaluating the Generalizability of Video Question Answering Models

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

This paper proposes a Video Graph Transformer (VGT) model for Video Quetion Answering (VideoQA). VGT's uniqueness are two-fold: 1) it designs a dynamic graph transformer module which encodes video by explicitly capturing the visual objects, their relations, and dynamics for complex spatio-temporal reasoning; and 2) it exploits disentangled video and text Transformers for relevance comparison between the video and text to perform QA, instead of entangled crossmodal Transformer for answer classification. Vision-text communication is done by additional cross-modal interaction modules. With more reasonable video encoding and QA solution, we show that VGT can achieve much better performances on VideoQA tasks that challenge dynamic relation reasoning than prior arts in the pretraining-free scenario. Its performances even surpass those models that are pretrained with millions of external data. We further show that VGT can also benefit a lot from selfsupervised cross-modal pretraining, yet with orders of magnitude smaller data. These results clearly demonstrate the effectiveness and superiority of VGT, and reveal its potential for more data-efficient pretraining. With comprehensive analyses and some heuristic observations, we hope that VGT can promote VQA research beyond coarse recognition/description towards fine-grained relation reasoning in realistic videos. Our code is available at https://github.com/sail-sg/VGT.

For agents

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

get_harvested_code_for_paper("2207.05342")
get_code_for_paper("2207.05342")
have("2207.05342")

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

curl https://syntology.ai/api/ran/2207.05342.json

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

[![Syntology run record](https://syntology.ai/api/ran/2207.05342.svg)](https://syntology.ai/paper/2207.05342)

Connect an agent — have() is free.