SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2301.00746 · CVPR · 2023

NaQ: Leveraging Narrations as Queries to Supervise Episodic Memory

U Austin, Kristen Grauman, Ziad Al-Halah, Santhosh Kumar

arXiv · PDF · Open in the Atlas

Code that ran

12 of 15 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 3 honoured the contract we drafted, 0 violated it and 5 have no contract verdict; 4 failures of our instrument; 3 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/2301.00746.json · badge: /api/ran/2301.00746.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
srama2512/NaQ canonical (classifier label)
classifier basis not recorded on the link
12 ran · 3 no recorded run
FunctionStatusWhere it lives
swish Ran
ran; our contract draft was wrong, not the code
srama2512/NaQ/ReLER/ms_cm/bert_layers.py
code served (permissive licence) · get_code("0f786c407fb1ee4c")
How it ran

Ran in Syntology's jail on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. The comparison behind this sample's level is a different run: the peer-agreement lane ran it again on the host, not in the jail (report dated 2026-09-04T22:33:52Z). 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)

# The function swish takes a single parameter 'x' which is used with:
# - torch.sigmoid(x) - requires a tensor
# - x * ... - requires a tensor that supports multiplication
# 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
count_parameters Ran
ran, honoured the contract we drafted
srama2512/NaQ/ReLER/utils/model_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from Zhuo-Cao/QV-M2, is pointer only (licence: NONE) · get_code("edb929dece68582e")
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.

filter_glove_embedding Ran
ran on a synthesized input
srama2512/NaQ/VSLNet/utils/data_gen.py
code served (permissive licence) · get_code("094900ebac5a22d2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (5, 300) 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 codecs
import numpy as np
import tempfile
import os

# Create a temporary GloVe-format file for testing
temp_dir = tempfile.mkdtemp()
glove_path = os.path.join(temp_dir, "glove.txt")

# Write a small test GloVe file with the correct format
# Format: word followed by 300 float values
with open(glove_path, 'w', encoding='utf-8') as f:
    # Write some valid entries (word + 300 values)
    f.write("the " + " ".join([str(0.1 * i) for i in range(300)]) + "\n")
    f.write("and " + " ".join([str(0.2 * i) for i in range(300)]) + "\n")
    f.write("of " + " ".join([str(0.3 * i) for i in range(300)]) + "\n")
    # Write an invalid entry (wrong number of values)
    f.write("invalid_word " + " ".join([str(0.1) for _ in range(50)]) + "\n")
    # Write another valid entry
    f.write("a " + " ".join([str(0.4 * i) for i in range(300)]) + "\n")

# Create a word_dict with indices for words that exist in the GloVe file
word_dict = {
    "the": 0,
    "and": 1,
    "of": 2,
    "a": 3,
    "unknown": 4
}

ARGS = {
    "word_dict": word_dict,
    "glove_path": glove_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (5, 300) — shape (5, 300) 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
gelu Ran
ran, honoured the contract we drafted
srama2512/NaQ/ReLER/ms_cm/bert_layers.py
pointer only (licence: MIT); the stored copy get_code serves, from maximek3/e-ViL, is pointer only (licence: NONE) · get_code("3c06d11eadebd504")
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 maximek3/e-ViL, 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.

get_em_clips Ran
ran on a synthesized input
srama2512/NaQ/utils/create_naq_dataset.py
code served (permissive licence) · get_code("e46b6ba1c089129f")
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 os
import json
import tempfile

# Create a temporary directory for annotations
temp_dir = tempfile.mkdtemp()

# Create mock annotation files for vq, nlq, and moments tasks
mock_data = {
    'videos': [
        {
            'video_uid': 'video_001',
            'clips': [
                {
                    'clip_uid': 'clip_001',
                    'video_start_sec': 0.0,
                    'video_end_sec': 5.0
                },
                {
                    'clip_uid': 'clip_002',
                    'video_start_sec': 5.0,
                    'video_end_sec': 10.0
                }
            ]
        },
        {
            'video_uid': 'video_002',
            'clips': [
                {
                    'clip_uid': 'clip_003',
                    'video_start_sec': 0.0,
                    'video_end_sec': 3.0
                }
            ]
        }
    ]
}

# Write mock annotation files for each task
for task in ['vq', 'nlq', 'moments']:
    file_path = os.path.join(temp_dir, f'{task}_train.json')
    with open(file_path, 'w') as f:
        json.dump(mock_data, f)
    
    # Also create v2 versions
    file_path_v2 = os.path.join(temp_dir, f'{task}_train_v2.json')
    with open(file_path_v2, 'w') as f:
        json.dump(mock_data, f)

ARGS = {
    'annotations_dir': temp_dir,
    'split': 'train',
    'em_version': 'v1'
}
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 — 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
load_glove Ran
ran on a synthesized input
srama2512/NaQ/VSLNet/utils/data_gen.py
code served (permissive licence) · get_code("623b22c7208a83a3")
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 set. 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 tempfile
import os

# Create a temporary GloVe file for testing
temp_dir = tempfile.mkdtemp()
glove_path = os.path.join(temp_dir, "glove.txt")

# Write a small sample GloVe file with proper formatting
# GloVe format: word followed by 300 dimensions
with open(glove_path, 'w', encoding='utf-8') as f:
    # Valid entries (word + 300 floats = 301 tokens when split)
    f.write("the " + " ".join(["0.1"] * 300) + "\n")
    f.write("a " + " ".join(["0.2"] * 300) + "\n")
    f.write("and " + " ".join(["0.3"] * 300) + "\n")
    # Invalid entries (should be skipped due to len(line) == 2 or len(line) != 301)
    f.write("invalid_short\n")
    f.write("bad " + " ".join(["0.4"] * 299) + "\n")

ARGS = {
    "glove_path": glove_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type set — a set output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a set 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
load_json Ran
ran; our contract draft was wrong, not the code
srama2512/NaQ/ReLER/ms_cm/vslnet_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from Alvin-Zeng/temporal-robustness-benchmark, is pointer only (licence: NONE) · get_code("a8ccf61e0ee23809")
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 Alvin-Zeng/temporal-robustness-benchmark, 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_json Ran
ran on a synthesized input
srama2512/NaQ/utils/create_naq_dataset.py
code served (permissive licence) · get_code("21e3f8885442c5f3")
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 for the function to load
temp_dir = tempfile.mkdtemp()
json_file = os.path.join(temp_dir, "test.json")

# Write test data to the JSON file
test_data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "key3": {"nested": "object"}
}

with open(json_file, 'w') as fp:
    json.dump(test_data, fp)

# Define arguments for the load_json function
ARGS = {
    "path": json_file
}
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
load_lines Ran
ran on a synthesized input
srama2512/NaQ/ReLER/ms_cm/vslnet_utils.py
code served (permissive licence) · get_code("f319d88d5afe29f2")
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 1, 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 tempfile
import os

# Create a temporary file with some test content
temp_file = tempfile.NamedTemporaryFile(mode='w', encoding='utf-8', delete=False, suffix='.txt')
temp_file.write("line one\n")
temp_file.write("line two\n")
temp_file.write("line three\n")
temp_file.close()

ARGS = {
    'filename': temp_file.name
}
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 1, dtype family float — 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
load_pickle Ran
ran; our contract draft was wrong, not the code
srama2512/NaQ/ReLER/ms_cm/vslnet_utils.py
code served (permissive licence) · get_code("259b0af842be3b00")
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 pickle
import tempfile
import os

# Create a temporary file with pickled data
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_data.pkl")

# Create some sample data to pickle
sample_data = {
    "key1": [1, 2, 3],
    "key2": {"nested": "value"},
    "key3": 42.5
}

# Write the pickled data to the temporary file
with open(temp_file, mode='wb') as handle:
    pickle.dump(sample_data, handle)

ARGS = {
    "filename": temp_file
}
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
mask_logits Ran
ran, honoured the contract we drafted
srama2512/NaQ/ReLER/ms_cm/model.py
pointer only (licence: MIT); the stored copy get_code serves, from mingyao1120/TR-DETR, is pointer only (licence: NOASSERTION) · get_code("60df4d5a1eee1b92")
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.

pad_text_tenor_list Ran
ran; our contract draft was wrong, not the code
srama2512/NaQ/ReLER/ms_cm/model.py
code served (permissive licence) · get_code("4789ccc84c2e515c")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 numpy as np

torch.manual_seed(0)

# Test case 1: List of 1D lists
sequences_1d = [[1, 2, 3], [1, 2], [3, 4, 7, 9]]

# Test case 2: List of 2D torch tensors with different first dimensions
sequences_2d = [
    torch.randn(2, 3),
    torch.randn(4, 3),
    torch.randn(1, 3)
]

# Test case 3: List of 3D torch tensors
sequences_3d = [
    torch.randn(2, 3, 4),
    torch.randn(4, 3, 4),
    torch.randn(1, 3, 4)
]

ARGS = {
    "sequences": sequences_3d,
    "dtype": torch.long,
    "device": torch.device('cpu'),
    "fixed_length": None
}
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 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
build_model Not yet run
unverified (harvested, no recorded run)
srama2512/NaQ/ReLER/ms_cm/clip_model.py
code served (permissive licence) · get_code("9312610a229b4ebf")
How it ran

no invocation recorded.

load_narrations_data Not yet run
unverified (harvested, no recorded run)
srama2512/NaQ/utils/create_naq_dataset.py
code served (permissive licence) · get_code("32e7ce1faa537e14")
How it ran

no invocation recorded.

vocab_emb_gen Not yet run
unverified (harvested, no recorded run)
srama2512/NaQ/VSLNet/utils/data_gen.py
code served (permissive licence) · get_code("76e886ecae69991a")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 12 are also present in other papers whose code we hold (12 with at least one paper dated earlier, 2 the same month, 7 later) and 3 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 206 other papers
128 dated earlier, 2 the same month, 76 later · 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 198 other papers
130 dated earlier, 2 the same month, 65 later, 1 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
build_model 119 other papers
23 dated earlier, 91 later, 3 same year, order not determinable, 2 undated
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1901.07441 (2019-01)
PadChest: A large chest x-ray image dataset with multi-label annotated reports
2104.14294 (2021-04)
Emerging Properties in Self-Supervised Vision Transformers
and 116 more not named here
count_parameters 11 other papers
3 dated earlier, 8 later
2001.09099 (2020-01)
TVR: A Large-Scale Dataset for Video-Subtitle Moment Retrieval
2105.06247 (2021-05)
Video Corpus Moment Retrieval with Contrastive Learning
2207.00383 (2022-07)
ReLER@ZJU-Alibaba Submission to the Ego4D Natural Language Queries Challenge 2022
and 8 more not named here
mask_logits 5 other papers
2 dated earlier, 3 later · 1 only as an assembly brick
2004.13931 (2020-04)
Span-based Localizing Network for Natural Language Video Localization
2207.00383 (2022-07)
ReLER@ZJU-Alibaba Submission to the Ego4D Natural Language Queries Challenge 2022
2307.16715 (2023-07)
UniVTG: Towards Unified Video-Language Temporal Grounding
and 2 more not named here
load_json 2 other papers
1 dated earlier, 1 later
2207.00383 (2022-07)
ReLER@ZJU-Alibaba Submission to the Ego4D Natural Language Queries Challenge 2022
2403.20254 (2024-03)
Benchmarking the Robustness of Temporal Action Detection Models Against Temporal Corruptions
load_pickle 2 other papers
1 dated earlier, 1 later
2207.00383 (2022-07)
ReLER@ZJU-Alibaba Submission to the Ego4D Natural Language Queries Challenge 2022
2406.00439 (2024-06)
Learning Manipulation by Predicting Interaction
filter_glove_embedding 1 other paper
1 dated earlier
2207.00383 (2022-07)
ReLER@ZJU-Alibaba Submission to the Ego4D Natural Language Queries Challenge 2022
load_glove 1 other paper
1 dated earlier
2207.00383 (2022-07)
ReLER@ZJU-Alibaba Submission to the Ego4D Natural Language Queries Challenge 2022
load_lines 1 other paper
1 dated earlier
2207.00383 (2022-07)
ReLER@ZJU-Alibaba Submission to the Ego4D Natural Language Queries Challenge 2022
pad_text_tenor_list 1 other paper
1 dated earlier
2207.00383 (2022-07)
ReLER@ZJU-Alibaba Submission to the Ego4D Natural Language Queries Challenge 2022
vocab_emb_gen 1 other paper
1 dated earlier
2207.00383 (2022-07)
ReLER@ZJU-Alibaba Submission to the Ego4D Natural Language Queries Challenge 2022

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

Searching long egocentric videos with natural language queries (NLQ) has compelling applications in augmented reality and robotics, where a fluid index into everything that a person (agent) has seen before could augment human memory and surface relevant information on demand. However, the structured nature of the learning problem (freeform text query inputs, localized video temporal window outputs) and its needle-in-a-haystack nature makes it both technically challenging and expensive to supervise. We introduce Narrations-as-Queries (NaQ), a data augmentation strategy that transforms standard video-text narrations into training data for a video query localization model. Validating our idea on the Ego4D benchmark, we find it has tremendous impact in practice. NaQ improves multiple top models by substantial margins (even doubling their accuracy), and yields the very best results to date on the Ego4D NLQ challenge, soundly outperforming all challenge winners in the CVPR and ECCV 2022 competitions and topping the current public leaderboard. Beyond achieving the state-of-the-art for NLQ, we also demonstrate unique properties of our approach such as the ability to perform zero-shot and few-shot NLQ, and improved performance on queries about long-tail object categories. Code and models: http://vision.cs.utexas.edu/projects/naq.

For agents

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

get_harvested_code_for_paper("2301.00746")
get_code_for_paper("2301.00746")
have("2301.00746")

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

curl https://syntology.ai/api/ran/2301.00746.json

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

[![Syntology run record](https://syntology.ai/api/ran/2301.00746.svg)](https://syntology.ai/paper/2301.00746)

Connect an agent — have() is free.