SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2503.10076 · ICCV · 2025

VMBench: A Benchmark for Perception-Aligned Video Motion Generation

Chen Zhu, Xiangxiang Chu, Group, Hangyu Li, Jiahong Wu, Jiashu Zhu, Xiaokun Feng, Meiqi Wu, Cundian Yang, Alibaba Amap, Aiming Hao, Xinran Ling, and 1 more

arXiv · PDF · Open in the Atlas

Code that ran

6 of 6 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 4 have no contract verdict; 1 failure of our instrument; 0 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/2503.10076.json · badge: /api/ran/2503.10076.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
AMAP-ML/VMBench canonical (classifier label)
classifier basis not recorded on the link
6 ran · 0 no recorded run
FunctionStatusWhere it lives
extract_frames_from_video Ran
ran on a synthesized input
AMAP-ML/VMBench/temporal_coherence_score.py
code served (permissive licence) · get_code("c5ea12db107359b3")
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 cv2
import numpy as np
import tempfile
import os

# Create a temporary video file for testing
temp_dir = tempfile.mkdtemp()
video_path = os.path.join(temp_dir, "test_video.mp4")

# Create a simple test video using OpenCV
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
fps = 30
frame_size = (640, 480)
out = cv2.VideoWriter(video_path, fourcc, fps, frame_size)

# Write a few test frames
for i in range(3):
    frame = np.zeros((480, 640, 3), dtype=np.uint8)
    frame[:, :, i % 3] = 100 + i * 50  # Different color for each frame
    out.write(frame)

out.release()

ARGS = {
    'video_path': video_path,
    'jpeg_quality': 95
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, cv2, numpy — 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
get_artifacts_frames Ran
ran on a synthesized input
AMAP-ML/VMBench/motion_smoothness_score.py
code served (permissive licence) · get_code("dccf6af3b73d6aad")
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 1, dtype family int, shape (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 numpy as np

# Create sample scores array with some variation
# to produce meaningful differences when np.diff is applied
scores = np.array([0.1, 0.15, 0.12, 0.5, 0.48, 0.2, 0.21, 0.19])

# Define threshold parameter
threshold = 0.025

ARGS = {
    "scores": scores,
    "threshold": threshold,
}
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 ndarray, ndim 1, dtype family int, shape (6,) — shape (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
get_loss_scale_for_deepspeed Ran
ran, honoured the contract we drafted
AMAP-ML/VMBench/VideoMAEv2/engine_for_finetuning.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from kangsankim07/videoicl, is pointer only (licence: NONE) · get_code("29416043c7035e7c")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from kangsankim07/videoicl, 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.

object_info_to_dict Ran
ran on a synthesized input
AMAP-ML/VMBench/temporal_coherence_score.py
code served (permissive licence) · get_code("07eeaf67c614f428")
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
import numpy as np

torch.manual_seed(0)

class ObjectInfo:
    def __init__(self, instance_id, mask, class_name, x1, y1, x2, y2, logit):
        self.instance_id = instance_id
        self.mask = mask
        self.class_name = class_name
        self.x1 = x1
        self.y1 = y1
        self.x2 = x2
        self.y2 = y2
        self.logit = logit

mask_tensor = torch.rand(4, 3)

obj = ObjectInfo(
    instance_id=42,
    mask=mask_tensor,
    class_name="person",
    x1=10,
    y1=20,
    x2=100,
    y2=150,
    logit=0.95
)

ARGS = {
    'obj': obj
}
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 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
set_threshold Ran
ran on a synthesized input
AMAP-ML/VMBench/motion_smoothness_score.py
code served (permissive licence) · get_code("6f8f7b238244ce38")
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 float, 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 torch
import numpy as np

torch.manual_seed(0)

# The function takes a single parameter 'camera_movement' which is compared against None and numeric thresholds
# It should be a numeric value (float or int) or None

ARGS = {
    "camera_movement": 0.2
}
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 float, 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
train_class_batch Ran
ran; our contract draft was wrong, not the code
AMAP-ML/VMBench/VideoMAEv2/engine_for_finetuning.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from kangsankim07/videoicl, is pointer only (licence: NONE) · get_code("c7976ea27edc377a")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from kangsankim07/videoicl, 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.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 0 later) and 4 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_loss_scale_for_deepspeed 15 other papers
15 dated earlier
2203.14415 (2022-03)
Mugs: A Multi-Granular Self-Supervised Learning Framework
2211.09120 (2022-11)
AdaMAE: Adaptive Masking for Efficient Spatiotemporal Learning with Masked Autoencoders
2212.10368 (2022-12)
Masked Event Modeling: Self-Supervised Pretraining for Event Cameras
and 12 more not named here
train_class_batch 11 other papers
11 dated earlier
2203.14415 (2022-03)
Mugs: A Multi-Granular Self-Supervised Learning Framework
2211.09120 (2022-11)
AdaMAE: Adaptive Masking for Efficient Spatiotemporal Learning with Masked Autoencoders
2212.10368 (2022-12)
Masked Event Modeling: Self-Supervised Pretraining for Event Cameras
and 8 more not named here

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 generation has advanced rapidly, improving evaluation methods, yet assessing video's motion remains a major challenge. Specifically, there are two key issues: 1) current motion metrics do not fully align with human perceptions; 2) the existing motion prompts are limited. Based on these findings, we introduce VMBench-a comprehensive Video Motion Benchmark that has perception-aligned motion metrics and features the most diverse types of motion. VMBench has several appealing properties: (1) Perception-Driven Motion Evaluation Metrics, we identify five dimensions based on human perception in motion video assessment and develop fine-grained evaluation metrics, providing deeper insights into models' strengths and weaknesses in motion quality. (2) Meta-Guided Motion Prompt Generation, a structured method that extracts meta-information, generates diverse motion prompts with LLMs, and refines them through human-AI validation, resulting in a multilevel prompt library covering six key dynamic scene dimensions. (3) Human-Aligned Validation Mechanism, we provide human preference annotations to validate our benchmarks, with our metrics achieving an average 35.3% improvement in Spearman's correlation over baseline methods. This is the first time that the quality of motion in videos has been evaluated from the perspective of human perception alignment. Additionally, we release VMBench at https://github.com/AMAP-ML/VMBench, setting a new standard for evaluating and advancing motion generation models.

For agents

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

get_harvested_code_for_paper("2503.10076")
get_code_for_paper("2503.10076")
have("2503.10076")

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

curl https://syntology.ai/api/ran/2503.10076.json

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

[![Syntology run record](https://syntology.ai/api/ran/2503.10076.svg)](https://syntology.ai/paper/2503.10076)

Connect an agent — have() is free.