SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2401.01529 · NeurIPS · 2023

Glance and Focus: Memory Prompting for Multi-Event Video Question Answering

Xilin Chen, Ziyi Bai, Ruiping Wang

arXiv · PDF · Open in the Atlas

Code that ran

14 of 18 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 2 honoured the contract we drafted, 1 violated it and 2 have no contract verdict; 9 failures of our instrument; 4 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/2401.01529.json · badge: /api/ran/2401.01529.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
ByZ0e/Glance-Focus canonical (classifier label)
classifier basis: census
12 ran · 3 no recorded run
byz0e/glance-focus canonical (classifier label)
classifier basis: classifier_2025
2 ran · 1 no recorded run
FunctionStatusWhere it lives
FCNet Ran
ran on a synthesized input · constructed an object; no result was computed
byz0e/glance-focus/model/glance_focus.py
code served (permissive licence) · get_code("d725dcf4cf20da7f")
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, FCNet: 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)

# The FCNet constructor takes:
# 1. dims: a list of integers representing layer dimensions
# 2. if_weight_norm: a boolean flag for whether to use weight normalization

ARGS = {
    "dims": [4, 8, 6, 3],  # Input dim 4, hidden dims 8 and 6, output dim 3
    "if_weight_norm": False
}
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 FCNet — 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 FCNet 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
GF Ran
ran on a synthesized input · constructed an object; no result was computed
byz0e/glance-focus/model/glance_focus.py
code served (permissive licence) · get_code("440bd21060030be0")
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, GF: 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 torch.nn.functional as F
from typing import Optional
from torch.nn.utils.weight_norm import weight_norm

torch.manual_seed(0)

class SimpleTransformer(nn.Module):
    """Minimal transformer module for testing"""
    def __init__(self, d_model=8):
        super().__init__()
        self.d_model = d_model
        self.encoder = nn.Linear(d_model, d_model)
        self.decoder = nn.Linear(d_model, d_model)
    
    def forward(self, src, mask):
        # Return encoded memory
        return self.encoder(src)

class FCNet(nn.Module):
    """Simple class for non-linear fully connect network"""
    def __init__(self, dims, if_weight_norm=False):
        super(FCNet, self).__init__()
        layers = []
        for i in range(len(dims) - 2):
            in_dim = dims[i]
            out_dim = dims[i + 1]
            if if_weight_norm:
                layers.append(weight_norm(nn.Linear(in_dim, out_dim), dim=None))
            else:
                layers.append(nn.Linear(in_dim, out_dim))
            layers.append(nn.ReLU())
        if if_weight_norm:
            layers.append(weight_norm(nn.Linear(dims[-2], dims[-1]), dim=None))
        else:
            layers.append(nn.Linear(dims[-2], dims[-1]))
        layers.append(nn.ReLU())
        self.main = nn.Sequential(*layers)

    def forward(self, x):
        return self.main(x)

class MLP(nn.Module):
    """ Very simple multi-layer perceptron (also called FFN)"""
    def __init__(self, input_dim, hidden_dim, output_dim, num_layers):
        super().__init__()
        self.num_layers = num_layers
        h = [hidden_dim] * (num_layers - 1)
        self.layers = nn.ModuleList((nn.Linear(n, k) for n, k in zip([input_dim] + h, h + [output_dim])))

    def forward(self, x):
        for i, layer in enumerate(self.layers):
            x = F.relu(layer(x)) if i < self.num_layers - 1 else layer(x)
        return x

class GF(nn.Module):
    """ This is the 2-stage Transformer Encoder-Decoder module """
    def __init__(self, transformer, num_queries, feature_dim, output_dim, event_pred_dim, qa_dataset: Optional[str]=None):
        super().__init__()
        self.num_queries = num_queries
        self.transformer = transformer
        self.output_dim = output_dim
        hidden_dim = transformer.d_model
        span_pred_dim = 2
        self.span_embed = MLP(hidden_dim, hidden_dim, span_pred_dim, 3)
        self.event_class_embed = nn.Linear(hidden_dim, event_pred_dim)
        self.query_embed = nn.Embedding(num_queries, hidden_dim)
        self.input_proj = FCNet([feature_dim, hidden_dim], if_weight_norm=True)
        self.position_embeddings = nn.Embedding(num_queries, hidden_dim)
        self.qa_embed = nn.Embedding(1, hidden_dim)
        self.qa_dataset = qa_dataset
        if qa_dataset is not None:
            if qa_dataset == 'star' or qa_dataset == 'nextqa':
                self.answer_head = nn.Sequential(nn.Linear(hidden_dim, hidden_dim * 2), nn.ReLU(True), nn.Linear(hidden_dim * 2, 1))
            else:
                self.answer_head = nn.Linear(hidden_dim, self.output_dim)

    def forward(self, src, mask, captions, encode_and_save=True, memory_cache=None, query_type=None, glance=None):
        pass

# Create minimal valid arguments
transformer = SimpleTransformer(d_model=8)
num_queries = 3
feature_dim = 6
output_dim = 5
event_pred_dim = 7
qa_dataset = None

# For forward() call
batch_size = 2
src_frame_features = torch.randn(batch_size, 4, feature_dim)
src_memory_prompt = torch.randn(batch_size, num_queries, 8)
src = (src_frame_features, src_memory_prompt)
mask = torch.zeros(batch_size, 4, dtype=torch.bool)
captions = ["caption1", "caption2"]

ARGS = {
    "transformer": transformer,
    "num_queries": num_queries,
    "feature_dim": feature_dim,
    "output_dim": output_dim,
    "event_pred_dim": event_pred_dim,
    "qa_dataset": qa_dataset,
}
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 GF — 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 GF 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
chunk_list Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ByZ0e/Glance-Focus/dataset/agqa.py
code served (permissive licence) · get_code("c61a34a3c9e0df91")
How it ran

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

random.seed(0)

ARGS = {
    "examples": [1, 2, 3, 4, 5],
    "chunk_size": 2,
    "pad_to_divisible": True,
}
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 list, ndim 2, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
collate_func Ran
ran; our contract draft was wrong, not the code
ByZ0e/Glance-Focus/dataset/egotaskqa.py
code served (permissive licence) · get_code("6941a4a46f03fa07")
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
import torch.nn as nn

torch.manual_seed(0)

# Create sample batch items
# Each item is a tuple: (answer_encode, appearance_feat, motion_feat, question, question_encode, reasoning_type, span, hoi)

batch_item_1 = (
    5,  # answer_encode: int
    torch.randn(4, 8),  # appearance_feat: [temporal_steps=4, feat_dim=8]
    torch.randn(4, 8),  # motion_feat: [temporal_steps=4, feat_dim=8]
    "What is happening?",  # question: str
    torch.tensor([1, 2, 3, 4, 5]),  # question_encode: token ids
    "temporal",  # reasoning_type: str
    [0, 2],  # span: list of indices
    [("person", "action")]  # hoi: list of tuples
)

batch_item_2 = (
    3,  # answer_encode: int
    torch.randn(4, 8),  # appearance_feat: [temporal_steps=4, feat_dim=8]
    torch.randn(4, 8),  # motion_feat: [temporal_steps=4, feat_dim=8]
    "Who is doing it?",  # question: str
    torch.tensor([2, 3]),  # question_encode: shorter sequence
    "object",  # reasoning_type: str
    [1, 3],  # span: list of indices
    [("object", "interaction")]  # hoi: list of tuples
)

batch = [batch_item_1, batch_item_2]

ARGS = {
    "batch": batch
}
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
flat_list_of_lists Ran
ran, honoured the contract we drafted
ByZ0e/Glance-Focus/dataset/nextqa.py
pointer only (licence: MIT); the stored copy get_code serves, from nangongningyi/gume, is pointer only (licence: NONE) · get_code("cef10190a6f8f7e9")
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 nangongningyi/gume, 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.

load_imgs Ran
ran; our contract draft was wrong, not the code
ByZ0e/Glance-Focus/utils/basic_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from RuipingL/Situat3DChange, is pointer only (licence: CC-BY-4.0) · get_code("1eee3af921bc5d88")
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 RuipingL/Situat3DChange, and CC-BY-4.0, as GitHub detects it on that repository, is not in the permissive set Syntology serves inline. 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.

load_json Ran
ran; our contract draft was wrong, not the code
ByZ0e/Glance-Focus/utils/basic_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from RuipingL/Situat3DChange, is pointer only (licence: CC-BY-4.0) · get_code("bace1acf5a748097")
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 RuipingL/Situat3DChange, and CC-BY-4.0, as GitHub detects it on that repository, is not in the permissive set Syntology serves inline. 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.

load_pickle Ran
ran; our contract draft was wrong, not the code
ByZ0e/Glance-Focus/utils/basic_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from RuipingL/Situat3DChange, is pointer only (licence: CC-BY-4.0) · get_code("7de9d66696a307eb")
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 RuipingL/Situat3DChange, and CC-BY-4.0, as GitHub detects it on that repository, is not in the permissive set Syntology serves inline. 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.

span_cxw_to_xx Ran
ran, violated the contract we drafted
ByZ0e/Glance-Focus/utils/span_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from mingyao1120/TR-DETR, is pointer only (licence: NOASSERTION) · get_code("b7639c2e8b83837b")
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 mingyao1120/TR-DETR, and GitHub could not classify the licence of that repository (NOASSERTION). 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.

span_xx_to_cxw Ran
ran; our contract draft was wrong, not the code
ByZ0e/Glance-Focus/utils/span_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from Zhuo-Cao/QV-M2, is pointer only (licence: NONE) · get_code("7114f025044af9b4")
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 Zhuo-Cao/QV-M2, 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.

span_xx_to_cxw Ran
ran; our contract draft was wrong, not the code
ByZ0e/Glance-Focus/dataset/egotaskqa.py
code served (permissive licence) · get_code("93d1fd5afccb1ea1")
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 float, shape (2, 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)

# Test case 1: 2D tensor of shape (#windows, 2)
xx_spans_2d = torch.tensor([[0.0, 1.0], [0.2, 0.4]])

# Test case 2: 3D tensor of shape (..., 2) for higher-dimensional input
xx_spans_3d = torch.tensor([[[0.0, 1.0], [0.2, 0.4]]])

ARGS = {
    "xx_spans": xx_spans_2d
}
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 float, shape (2, 2) — shape (2, 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
temporal_iou Ran
ran; our contract draft was wrong, not the code
ByZ0e/Glance-Focus/utils/span_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from Zhuo-Cao/QV-M2, is pointer only (licence: NONE) · get_code("e4ae5582a2533cdb")
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 Zhuo-Cao/QV-M2, 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.

tokenize Ran
ran, honoured the contract we drafted
ByZ0e/Glance-Focus/dataset/nextqa.py
code served (permissive licence) · get_code("c31fe80a1b2ba813")
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, 10) 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)

class MockTokenizer:
    def batch_encode_plus(self, sequences, add_special_tokens=True, max_length=10, padding='longest', truncation=True):
        """Mock tokenizer that returns input_ids"""
        # Return a dict with 'input_ids' key containing token lists
        result = []
        for seq in sequences:
            # Simple mock: convert text length to token count
            tokens = [101] + [1000 + i for i in range(min(len(seq.split()), max_length - 2))] + [102]
            # Pad or truncate to max_length
            if len(tokens) < max_length and padding == 'longest':
                tokens = tokens + [0] * (max_length - len(tokens))
            elif len(tokens) > max_length:
                tokens = tokens[:max_length]
            else:
                tokens = tokens + [0] * (max_length - len(tokens))
            result.append(tokens)
        return {'input_ids': result}

seq = ["hello world", "foo bar baz"]
tokenizer = MockTokenizer()
add_special_tokens = True
max_length = 10
dynamic_padding = True
truncation = True

ARGS = {
    'seq': seq,
    'tokenizer': tokenizer,
    'add_special_tokens': add_special_tokens,
    'max_length': max_length,
    'dynamic_padding': dynamic_padding,
    'truncation': truncation
}
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, 10) — shape (2, 10) 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
videoqa_collate_fn Ran
ran; our contract draft was wrong, not the code
ByZ0e/Glance-Focus/dataset/nextqa.py
code served (permissive licence) · get_code("bab1acae729c21cf")
How it ran

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

import torch

torch.manual_seed(0)

# Construct a batch of dataset items
# Each item is a dict with: question, answer_txt, question_txt, video_id, question_id, video, answer_id, answer
batch = [
    {
        'question': torch.tensor([1, 2, 3], dtype=torch.long),
        'answer_txt': ['option1', 'option2'],
        'question_txt': ['what is this'],
        'video_id': 'vid_001',
        'question_id': 'q_001',
        'video': torch.randn(4, 4),
        'answer_id': 0,
        'answer': torch.tensor([[1, 2, 3], [4, 5, 6]], dtype=torch.long)
    },
    {
        'question': torch.tensor([7, 8], dtype=torch.long),
        'answer_txt': ['option3'],
        'question_txt': ['where is that'],
        'video_id': 'vid_002',
        'question_id': 'q_002',
        'video': torch.randn(4, 4),
        'answer_id': 1,
        'answer': torch.tensor([[10, 11], [12, 13]], dtype=torch.long)
    }
]

ARGS = {
    'batch': batch
}
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 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
build_transformer Not yet run
unverified (harvested, no recorded run)
ByZ0e/Glance-Focus/model/transformer_gf.py
code served (permissive licence) · get_code("e66b5e8b24dfea32")
How it ran

no invocation recorded.

mk_input_group Not yet run
unverified (harvested, no recorded run)
ByZ0e/Glance-Focus/dataset/agqa.py
code served (permissive licence) · get_code("6381e6df15302a0e")
How it ran

no invocation recorded.

repeat_tensor_rows Not yet run
unverified (harvested, no recorded run)
ByZ0e/Glance-Focus/dataset/agqa.py
code served (permissive licence) · get_code("bffb542b7e13c786")
How it ran

no invocation recorded.

sorter Not yet run
unverified (harvested, no recorded run)
byz0e/glance-focus/train_glance_focus_sup.py
code served (permissive licence) · get_code("0b0c5cf51bc1e61f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 distinct functions with an identity, 7 are also present in other papers whose code we hold (7 with at least one paper dated earlier, 3 the same month, 7 later) and 11 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
span_cxw_to_xx 10 other papers
3 dated earlier, 1 the same month, 6 later · 2 only as an assembly brick
2107.09609 (2021-07)
QVHIGHLIGHTS: Detecting Moments and Highlights in Videos via Natural Language Queries
2307.02869 (2023-07)
MomentDiff: Generative Video Moment Retrieval from Random to Real
2311.08835 (2023-11)
Correlation-Guided Query-Dependency Calibration for Video Temporal Grounding
and 7 more not named here
span_xx_to_cxw 9 other papers
3 dated earlier, 1 the same month, 5 later
2307.02869 (2023-07)
MomentDiff: Generative Video Moment Retrieval from Random to Real
2311.08835 (2023-11)
Correlation-Guided Query-Dependency Calibration for Video Temporal Grounding
2312.00083 (2023-12)
BAM-DETR: Boundary-Aligned Moment Detection Transformer for Temporal Sentence Grounding in Videos
and 6 more not named here
temporal_iou 8 other papers
2 dated earlier, 1 the same month, 5 later
2307.02869 (2023-07)
MomentDiff: Generative Video Moment Retrieval from Random to Real
2311.08835 (2023-11)
Correlation-Guided Query-Dependency Calibration for Video Temporal Grounding
2401.02309 (2024-01)
TR-DETR: Task-Reciprocal Transformer for Joint Moment Retrieval and Highlight Detection
and 5 more not named here
flat_list_of_lists 3 other papers
2 dated earlier, 1 later
1904.11574 (2019-04)
TVQA+: Spatio-Temporal Grounding for Video Question Answering
2001.09099 (2020-01)
TVR: A Large-Scale Dataset for Video-Subtitle Moment Retrieval
2407.12338 (2024-07)
GUME: Graphs and User Modalities Enhancement for Long-Tail Multimodal Recommendation
load_imgs 2 other papers
1 dated earlier, 1 later
2311.12871 (2023-11)
An Embodied Generalist Agent in 3D World
2510.11509 (2025-10)
Situat3DChange: Situated 3D Change Understanding Dataset for Multimodal Large Language Model
load_json 2 other papers
1 dated earlier, 1 later
2311.12871 (2023-11)
An Embodied Generalist Agent in 3D World
2510.11509 (2025-10)
Situat3DChange: Situated 3D Change Understanding Dataset for Multimodal Large Language Model
load_pickle 2 other papers
1 dated earlier, 1 later
2311.12871 (2023-11)
An Embodied Generalist Agent in 3D World
2510.11509 (2025-10)
Situat3DChange: Situated 3D Change Understanding Dataset for Multimodal Large Language Model

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

Video Question Answering (VideoQA) has emerged as a vital tool to evaluate agents' ability to understand human daily behaviors. Despite the recent success of large vision language models in many multi-modal tasks, complex situation reasoning over videos involving multiple human-object interaction events still remains challenging. In contrast, humans can easily tackle it by using a series of episode memories as anchors to quickly locate question-related key moments for reasoning. To mimic this effective reasoning strategy, we propose the Glance-Focus model. One simple way is to apply an action detection model to predict a set of actions as key memories. However, these actions within a closed set vocabulary are hard to generalize to various video domains. Instead of that, we train an Encoder-Decoder to generate a set of dynamic event memories at the glancing stage. Apart from using supervised bipartite matching to obtain the event memories, we further design an unsupervised memory generation method to get rid of dependence on event annotations. Next, at the focusing stage, these event memories act as a bridge to establish the correlation between the questions with high-level event concepts and low-level lengthy video content. Given the question, the model first focuses on the generated key event memory, then focuses on the most relevant moment for reasoning through our designed multi-level crossattention mechanism. We conduct extensive experiments on four Multi-Event VideoQA benchmarks including STAR, EgoTaskQA, AGQA, and NExT-QA. Our proposed model achieves state-of-the-art results, surpassing current large models in various challenging reasoning tasks. The code and models are available at https://github.com/ByZ0e/Glance-Focus. Recently, the emerged large vision-language models [1,3,8,25,47,50] have demonstrated impressive generalization capabilities in VideoQA tasks. By leveraging pre-training on large-scale cross-modal pairs, these models can establish correlations between questions and videos to find answers effectively. Such approaches proves effective for short videos containing single or few events with one-step 37th Conference on Neural Information Processing Systems (NeurIPS 2023).

For agents

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

get_harvested_code_for_paper("2401.01529")
get_code_for_paper("2401.01529")
have("2401.01529")

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

curl https://syntology.ai/api/ran/2401.01529.json

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

[![Syntology run record](https://syntology.ai/api/ran/2401.01529.svg)](https://syntology.ai/paper/2401.01529)

Connect an agent — have() is free.