SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1609.08675 · 2016

YouTube-8M: A Large-Scale Video Classification Benchmark

arXiv · PDF · Open in the Atlas

Code that ran

7 of 9 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; 1 honoured the contract we drafted, 0 violated it and 5 have no contract verdict; 1 failure of our instrument; 2 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/1609.08675.json · badge: /api/ran/1609.08675.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
google/youtube-8m canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
boseaslcohort/youtube-8m community (classifier label)
classifier basis not recorded on the link
5 ran · 2 no recorded run
taufikxu/youtube community (classifier label)
classifier basis not recorded on the link
1 ran · 0 no recorded run
FunctionStatusWhere it lives
FramePooling Ran
ran on a synthesized input
boseaslcohort/youtube-8m/model_utils.py
code served (permissive licence) · get_code("2f3219b9915649d3")
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 SymbolicTensor. 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 tensorflow as tf
import numpy as np

# Create a small tensor with shape [batch_size, num_frames, feature_size]
# batch_size=2, num_frames=3, feature_size=4
frames = tf.constant(np.random.randn(2, 3, 4).astype(np.float32))

# Test with "average" pooling method
method = "average"

ARGS = {
    "frames": frames,
    "method": method,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SymbolicTensor — a SymbolicTensor 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
calculate_hit_at_one Ran
ran on a synthesized input
boseaslcohort/youtube-8m/eval_util.py
code served (permissive licence) · get_code("427f933cdb30be37")
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 float32, ndim 0, dtype family float, shape () 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 numpy

# Set seed for reproducibility
numpy.random.seed(0)

# Create small numpy arrays for predictions and actuals
# predictions: batch_size x num_classes (2 x 4)
# Each row should contain model outputs (logits/scores) for each class
predictions = numpy.array([
    [0.1, 0.9, 0.5, 0.3],
    [0.2, 0.3, 0.8, 0.1]
], dtype=numpy.float32)

# actuals: batch_size x num_classes (2 x 4)
# Ground truth labels - should be binary (0 or 1) indicating correct class
# The function uses argmax on predictions, then indexes into actuals
# So actuals[batch_idx, predicted_class] should be 1 if correct, 0 if wrong
actuals = numpy.array([
    [0.0, 1.0, 0.0, 0.0],  # correct class is 1
    [0.0, 0.0, 1.0, 0.0]   # correct class is 2
], dtype=numpy.float32)

ARGS = {
    "predictions": predictions,
    "actuals": actuals
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float32, ndim 0, dtype family float, shape () — shape () 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
calculate_precision_at_equal_recall_rate Ran
ran on a synthesized input
boseaslcohort/youtube-8m/eval_util.py
code served (permissive licence) · get_code("a3d2406f6ce64e29")
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 float32, ndim 0, dtype family float, shape () 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 numpy

# Set seed for reproducibility
numpy.random.seed(0)

# Create sample predictions matrix (batch=2, num_classes=4)
# Values should be between 0 and 1 as model outputs
predictions = numpy.array([
    [0.9, 0.1, 0.8, 0.2],
    [0.3, 0.7, 0.4, 0.6]
], dtype=numpy.float32)

# Create sample actuals matrix (batch=2, num_classes=4)
# Binary labels indicating which classes are relevant
actuals = numpy.array([
    [1, 0, 1, 0],
    [0, 1, 0, 1]
], dtype=numpy.float32)

ARGS = {
    "predictions": predictions,
    "actuals": actuals,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float32, ndim 0, dtype family float, shape () — shape () 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
flatten Ran
ran, honoured the contract we drafted
boseaslcohort/youtube-8m/eval_util.py
code served (permissive licence) · get_code("557deb04a3b96763")
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 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 torch

torch.manual_seed(0)

# The function 'flatten' takes a list of lists and merges them into a single list
# It iterates through sublists and extracts individual items
# The parameter 'l' should be a list of lists

ARGS = {
    "l": [
        [1, 2, 3],
        [4, 5],
        [6, 7, 8, 9]
    ]
}
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 list, ndim 1, 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
get_segments Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
google/youtube-8m/inference.py
code served (permissive licence) · get_code("2ef42071ff785910")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) 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 numpy as np

np.random.seed(0)

# Create a small batch of video features
video_batch_size = 2
max_frame = 12
feature_dim = 4
segment_size = 3

# batch_video_mtx: (batch_size, max_frames, feature_dim)
batch_video_mtx = np.random.randn(video_batch_size, max_frame, feature_dim).astype(np.float32)

# batch_num_frames: (batch_size,) - actual number of frames per video
batch_num_frames = np.array([10, 8])

ARGS = {
    'batch_video_mtx': batch_video_mtx,
    'batch_num_frames': batch_num_frames,
    'segment_size': segment_size
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
numpy — 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
resize_axis Ran
ran on a synthesized input
taufikxu/youtube/readers.py
code served (permissive licence) · get_code("a8936fa247d02451")
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 EagerTensor, ndim 2, dtype family float, shape (3, 6) 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 tensorflow as tf

# Create a simple tensor to resize
tensor = tf.constant([[1.0, 2.0, 3.0, 4.0],
                      [5.0, 6.0, 7.0, 8.0],
                      [9.0, 10.0, 11.0, 12.0]])  # Shape: (3, 4)

# Define arguments for resize_axis
ARGS = {
    'tensor': tensor,
    'axis': 1,
    'new_size': 6,
    'fill_value': 0
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 2, dtype family float, shape (3, 6) — shape (3, 6) 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
to_csv_row Ran
ran on a synthesized input
boseaslcohort/youtube-8m/convert_prediction_from_json_to_csv.py
code served (permissive licence) · get_code("03117d2e243f0f99")
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 str. 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 numpy as np

torch.manual_seed(0)

# Construct valid arguments for to_csv_row
# The function expects json_data with 'video_id', 'class_indexes', 'predictions'
# video_id can be bytes or a list of bytes
# class_indexes and predictions are lists (or lists of lists) with equal length

json_data = {
    'video_id': b'video_123',
    'class_indexes': [0, 1, 2],
    'predictions': [0.95, 0.87, 0.72]
}

ARGS = {
    'json_data': json_data
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
SampleRandomFrames Not yet run
unverified (harvested, no recorded run)
boseaslcohort/youtube-8m/model_utils.py
code served (permissive licence) · get_code("c7c286e6493b9cf6")
How it ran

no invocation recorded.

SampleRandomSequence Not yet run
unverified (harvested, no recorded run)
boseaslcohort/youtube-8m/model_utils.py
code served (permissive licence) · get_code("604001c4ca602f10")
How it ran

no invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 distinct functions with an identity, 7 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 7 later) and 2 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
flatten 21 other papers
21 later
1802.04821 (2018-02)
Evolved Policy Gradients
1811.05014 (2018-11)
NeXtVLAD: An Efficient Neural Network to Aggregate Frame-level Features for Large-scale Video Classification
1911.03070 (2019-11)
Interactive Refinement of Cross-Lingual Word Embeddings
and 18 more not named here
FramePooling 1 other paper
1 later
1811.05014 (2018-11)
NeXtVLAD: An Efficient Neural Network to Aggregate Frame-level Features for Large-scale Video Classification
SampleRandomFrames 1 other paper
1 later
1811.05014 (2018-11)
NeXtVLAD: An Efficient Neural Network to Aggregate Frame-level Features for Large-scale Video Classification
SampleRandomSequence 1 other paper
1 later
1811.05014 (2018-11)
NeXtVLAD: An Efficient Neural Network to Aggregate Frame-level Features for Large-scale Video Classification
calculate_hit_at_one 1 other paper
1 later
1811.05014 (2018-11)
NeXtVLAD: An Efficient Neural Network to Aggregate Frame-level Features for Large-scale Video Classification
calculate_precision_at_equal_recall_rate 1 other paper
1 later
1811.05014 (2018-11)
NeXtVLAD: An Efficient Neural Network to Aggregate Frame-level Features for Large-scale Video Classification
to_csv_row 1 other paper
1 later
1811.05014 (2018-11)
NeXtVLAD: An Efficient Neural Network to Aggregate Frame-level Features for Large-scale Video Classification

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

Many recent advancements in Computer Vision are attributed to large datasets. Open-source software packages for Machine Learning and inexpensive commodity hardware have reduced the barrier of entry for exploring novel approaches at scale. It is possible to train models over millions of examples within a few days. Although large-scale datasets exist for image understanding, such as ImageNet, there are no comparable size video classification datasets. In this paper, we introduce YouTube-8M, the largest multi-label video classification dataset, composed of ~8 million videos (500K hours of video), annotated with a vocabulary of 4800 visual entities. To get the videos and their labels, we used a YouTube video annotation system, which labels videos with their main topics. While the labels are machine-generated, they have high-precision and are derived from a variety of human-based signals including metadata and query click signals. We filtered the video labels (Knowledge Graph entities) using both automated and manual curation strategies, including asking human raters if the labels are visually recognizable. Then, we decoded each video at one-frame-per-second, and used a Deep CNN pre-trained on ImageNet to extract the hidden representation immediately prior to the classification layer. Finally, we compressed the frame features and make both the features and video-level labels available for download. We trained various (modest) classification models on the dataset, evaluated them using popular evaluation metrics, and report them as baselines. Despite the size of the dataset, some of our models train to convergence in less than a day on a single machine using TensorFlow. We plan to release code for training a TensorFlow model and for computing metrics.

For agents

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

get_harvested_code_for_paper("1609.08675")
get_code_for_paper("1609.08675")
have("1609.08675")

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

curl https://syntology.ai/api/ran/1609.08675.json

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

[![Syntology run record](https://syntology.ai/api/ran/1609.08675.svg)](https://syntology.ai/paper/1609.08675)

Connect an agent — have() is free.