Papers › MVBench: A Comprehensive Multi-modal Video Understanding Benchmark

MVBench: A Comprehensive Multi-modal Video Understanding Benchmark

28 Nov 2023CVPR 2024 1arXiv:2311.17005archive 2025-07-28

Kunchang Li, Yali Wang, Yinan He, Yizhuo Li, Yi Wang, Yi Liu, Zun Wang, Jilan Xu, Guo Chen, Ping Luo, LiMin Wang, Yu Qiao

With the rapid development of Multi-modal Large Language Models (MLLMs), a number of diagnostic benchmarks have recently emerged to evaluate the comprehension capabilities of these models. However, most benchmarks predominantly assess spatial understanding in the static image tasks, while overlooking temporal understanding in the dynamic video tasks. To alleviate this issue, we introduce a comprehensive Multi-modal Video understanding Benchmark, namely MVBench, which covers 20 challenging video tasks that cannot be effectively solved with a single frame. Specifically, we first introduce a novel static-to-dynamic method to define these temporal-related tasks. By transforming various static tasks into dynamic ones, we enable the systematic generation of video tasks that require a broad spectrum of temporal skills, ranging from perception to cognition. Then, guided by the task definition, we automatically convert public video annotations into multiple-choice QA to evaluate each task. On one hand, such a distinct paradigm allows us to build MVBench efficiently, without much manual intervention. On the other hand, it guarantees evaluation fairness with ground-truth video annotations, avoiding the biased scoring of LLMs. Moreover, we further develop a robust video MLLM baseline, i.e., VideoChat2, by progressive multi-modal training with diverse instruction-tuning data. The extensive results on our MVBench reveal that, the existing MLLMs are far from satisfactory in temporal understanding, while our VideoChat2 largely surpasses these leading models by over 15% on MVBench. All models and data are available at https://github.com/OpenGVLab/Ask-Anything.

PaperPDFConference PDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 10 code samples harvested from 1 repository linked to this paper: 7 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 1 violated, 2 with no contract checked; 4 where Syntology's instrument failed) · 3 unverified. Where Syntology's instrument failed: 2 where our contract draft was wrong; 2 where the contract check could not run; failures of Syntology's instrument, not of the code.

By repository: official repository: 10 samples from 1 repository, 7 ran (3 with no instrument failure; 4 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

No sample is linked to this paper from 2 of its repositories that Syntology harvested for other papers:

Levels written between 2026-09-22 and 2026-09-24 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed. 4 of the 7 samples that ran have no recorded date.

opengvlab/ask-anything officialmentioned in papermentioned on GitHubpytorchMITSyntology: 10 samples, 7 ran (4 instrument-failed) · 3 unverified report
bytedance/tarsier mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Tarsier: Recipes for Training and Evaluating Large Video Description Models (arXiv:2407.00634; that paper's own run record: 2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report
magic-research/PLLaVA mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to PLLaVA : Parameter-free LLaVA Extension from Images to Videos for Video Dense Captioning (arXiv:2404.16994; that paper's own run record: 0 ran · 2 unverified); nothing checks that they implement this paper's method. report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

10 samples harvested: 7 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 1 violated, 2 with no contract checked; 4 where Syntology's instrument failed) · 3 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

1ran · violated contract
2ran · our draft was wrong
2ran; the contract check could not run
2ran
3unverified

Licence: 2 of the 10 samples are pointer only, meaning Syntology does not serve that copy's text. For 2 of them, this paper's copy is under a permissive licence and cleared, but the stored copy of the same code that Syntology serves is not; the licence label's tooltip gives Syntology's recorded reason. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from opengvlab/ask-anything. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.

opengvlab/ask-anything official repository 10 samples, 7 ran (4 instrument-failed) · 3 unverified
apply_rotary_pos_emb in 393 papers
census: 14 implementations from 15 papers, 8 distinct outputs across 7 buckets, one shared input per bucket
video_chat/models/modeling_llama.py ran; the contract check could not run (RAISES) · metamorphic tier: invariant MIT (permissive) · 9b4dff79d5e6102c · report fixture: signature q:Tensor/4/float, k:Tensor/4/float, cos:Tensor/4/float, sin:Tensor/4/float, position_ids:Tensor/2/int · returned tuple · level recorded, date not recorded
disabled_train in 172 papers video_chat/models/blip2.py ran · violated contract MIT (permissive) · 4cb732f513d69dfd · report fixture: signature self:callable/0/?, mode:bool/0/bool · returned SimpleModel · level recorded, date not recorded
get_prompt in 99 papers video_chat/conversation.py ran · our draft was wrong MIT (permissive) · 748d5a2586ca3178 · report fixture: signature conv:Conv/0/? · returned str · level recorded, date not recorded
get_sim in 12 papers video_chat2/models/criterions.py ran MIT (permissive) here; the stored copy is pointer only · a475c4129bb51880 · report fixture: signature vision_proj:Tensor/3/float, text_proj:Tensor/2/float, temp:float/0/float, agg_method:str/0/str · returned tuple · level recorded 2026-09-24
interpolate_temporal_pos_embed in 6 papers video_chat2/models/utils.py ran; the contract check could not run (RAISES) · metamorphic tier: invariant MIT (permissive) · f64c56c357505158 · report fixture: signature temp_embed_old:Tensor/4/float, num_frames_new:int/0/int · returned Tensor · level recorded 2026-09-22
load_temp_embed_with_mismatch in 5 papers video_chat2/models/utils.py ran fingerprinted MIT (permissive) here; the stored copy is pointer only · 5349fd1148712a53 · report fixture: signature temp_embed_old:Tensor/4/float, temp_embed_new:Tensor/4/float, add_zero:bool/0/bool · returned Tensor · level recorded 2026-09-24
rotate_half in 424 papers
census: 24 implementations from 96 papers, 4 distinct outputs across 2 buckets, one shared input per bucket this sample: bucket 1 of 2, arg 1: rank 2, kind float, dtype float32
video_chat/models/modeling_llama.py ran · our draft was wrong fingerprinted MIT (permissive) · b99eea6376d1e212 · report fixture: signature x:Tensor/2/float · returned Tensor · level recorded, date not recorded
create_eva_vit_g in 13 papers video_chat/models/eva_vit.py unverified MIT (permissive) · 19a98ad1eb25acff · report
inflate_weight video_chat/models/eva_vit.py unverified MIT (permissive) · bdec7f112532de68 · report
load_state_dict in 114 papers video_chat/models/eva_vit.py unverified MIT (permissive) · 0ba0ac7553b4501c · report

Tasks

3D Question Answering (3D-QA)DiagnosticFairnessMultiple-choiceVCGBench-DiverseVideo Question AnsweringVideo UnderstandingVideo-based Generative Performance BenchmarkingVideo-based Generative Performance Benchmarking (Consistency)Video-based Generative Performance Benchmarking (Contextual Understanding)Video-based Generative Performance Benchmarking (Correctness of Information)Video-based Generative Performance Benchmarking (Detail Orientation))Video-based Generative Performance Benchmarking (Temporal Understanding)Zero-Shot LearningZero-Shot Video Question Answer

1 archive task tag without a task page not shown.

Datasets

Introduced by this paper, per the archive.

MVBench

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
3D Question Answering (3D-QA) SQA3D VideoChat2 Exact Match 37.3 #12 of 13 Archive leaderboard report
3D Question Answering (3D-QA) ScanQA Test w/ objects VideoChat2 BLEU-4 9.6 #13 of 18 Archive leaderboard report
3D Question Answering (3D-QA) ScanQA Test w/ objects VideoChat2 CIDEr 49.2 #13 of 18 Archive leaderboard report
3D Question Answering (3D-QA) ScanQA Test w/ objects VideoChat2 Exact Match 19.2 #13 of 18 Archive leaderboard report
3D Question Answering (3D-QA) ScanQA Test w/ objects VideoChat2 METEOR 9.5 #13 of 18 Archive leaderboard report
3D Question Answering (3D-QA) ScanQA Test w/ objects VideoChat2 ROUGE 28.2 #13 of 18 Archive leaderboard report
VCGBench-Diverse VideoInstruct VideoChat2 Consistency 2.27 #3 of 6 Archive leaderboard report
VCGBench-Diverse VideoInstruct VideoChat2 Contextual Understanding 2.51 #3 of 6 Archive leaderboard report
VCGBench-Diverse VideoInstruct VideoChat2 Correctness of Information 2.13 #3 of 6 Archive leaderboard report
VCGBench-Diverse VideoInstruct VideoChat2 Dense Captioning 1.26 #3 of 6 Archive leaderboard report
VCGBench-Diverse VideoInstruct VideoChat2 Detail Orientation 2.42 #3 of 6 Archive leaderboard report
VCGBench-Diverse VideoInstruct VideoChat2 Reasoning 3.13 #3 of 6 Archive leaderboard report
VCGBench-Diverse VideoInstruct VideoChat2 Spatial Understanding 2.43 #3 of 6 Archive leaderboard report
VCGBench-Diverse VideoInstruct VideoChat2 Temporal Understanding 1.66 #3 of 6 Archive leaderboard report
VCGBench-Diverse VideoInstruct VideoChat2 mean 2.20 #3 of 6 Archive leaderboard report
Video Question Answering ActivityNet-QA VideoChat2 Accuracy 49.1 #8 of 36 Archive leaderboard report
Video Question Answering ActivityNet-QA VideoChat2 Confidence score 3.3 #8 of 36 Archive leaderboard report
Video Question Answering IntentQA VideoChat2_HD_mistral Accuarcy 83.4 #1 of 6 Archive leaderboard report
Video Question Answering IntentQA VideoChat2_HD_mistral CH 90.0 #1 of 6 Archive leaderboard report
Video Question Answering IntentQA VideoChat2_HD_mistral CW 84.0 #1 of 6 Archive leaderboard report
Video Question Answering IntentQA VideoChat2_HD_mistral TP&TN 77.3 #1 of 6 Archive leaderboard report
Video Question Answering IntentQA VideoChat2_mistral Accuarcy 81.9 #2 of 6 Archive leaderboard report
Video Question Answering IntentQA VideoChat2_mistral CH 86.9 #2 of 6 Archive leaderboard report
Video Question Answering IntentQA VideoChat2_mistral CW 82.6 #2 of 6 Archive leaderboard report
Video Question Answering IntentQA VideoChat2_mistral TP&TN 77.0 #2 of 6 Archive leaderboard report
Video Question Answering MVBench VideoChat2 Avg. 51.9 #13 of 22 Archive leaderboard report
Video Question Answering NExT-QA VideoChat2_HD_mistral Accuracy 79.5 #14 of 47 Archive leaderboard report
Video Question Answering NExT-QA VideoChat2_mistral Accuracy 78.6 #17 of 47 Archive leaderboard report
Video Question Answering NExT-QA VideoChat2 Accuracy 68.6 #31 of 47 Archive leaderboard report
Video Question Answering TVBench VideoChat2 Average Accuracy 35.0 #27 of 28 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2_HD_mistral Consistency 2.84 #10 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2_HD_mistral Contextual Understanding 3.72 #10 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2_HD_mistral Correctness of Information 3.40 #10 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2_HD_mistral Detail Orientation 2.91 #10 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2_HD_mistral Temporal Understanding 2.65 #10 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2_HD_mistral mean 3.10 #10 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2 Consistency 2.81 #15 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2 Contextual Understanding 3.51 #15 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2 Correctness of Information 3.02 #15 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2 Detail Orientation 2.88 #15 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2 Temporal Understanding 2.66 #15 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking VideoInstruct VideoChat2 mean 2.98 #15 of 23 Archive leaderboard report
Video-based Generative Performance Benchmarking (Consistency) VideoInstruct VideoChat2 gpt-score 2.81 #7 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Consistency) VideoInstruct VideoChat2_HD_mistral gpt-score 2.62 #10 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Contextual Understanding) VideoInstruct VideoChat2_HD_mistral gpt-score 3.64 #7 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Contextual Understanding) VideoInstruct VideoChat2 gpt-score 3.51 #9 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Correctness of Information) VideoInstruct VideoChat2_HD_mistral gpt-score 3.40 #5 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Correctness of Information) VideoInstruct VideoChat2 gpt-score 3.02 #9 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Detail Orientation)) VideoInstruct VideoChat2 gpt-score 2.88 #11 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Detail Orientation)) VideoInstruct VideoChat2_HD_mistral gpt-score 2.86 #12 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Temporal Understanding) VideoInstruct VideoChat2 gpt-score 2.66 #7 of 18 Archive leaderboard report
Video-based Generative Performance Benchmarking (Temporal Understanding) VideoInstruct VideoChat2_HD_mistral gpt-score 2.65 #9 of 18 Archive leaderboard report
Zero-Shot Learning TVQA VideoChat2 Accuracy 40.6 #1 of 1 Archive leaderboard report
Zero-Shot Video Question Answer ActivityNet-QA VideoChat2 Accuracy 49.1 #14 of 28 Archive leaderboard report
Zero-Shot Video Question Answer ActivityNet-QA VideoChat2 Confidence Score 3.3 #14 of 28 Archive leaderboard report
Zero-Shot Video Question Answer EgoSchema (fullset) VideoChat2_phi3 Accuracy 56.7 #12 of 29 Archive leaderboard report
Zero-Shot Video Question Answer EgoSchema (fullset) VideoChat2_HD_mistral Accuracy 55.8 #13 of 29 Archive leaderboard report
Zero-Shot Video Question Answer EgoSchema (fullset) VideoChat2_mistral Accuracy 54.4 #14 of 29 Archive leaderboard report
Zero-Shot Video Question Answer EgoSchema (subset) VideoChat2_HD_mistral Accuracy 65.6 #6 of 14 Archive leaderboard report
Zero-Shot Video Question Answer EgoSchema (subset) VideoChat2_mistral Accuracy 63.6 #7 of 14 Archive leaderboard report
Zero-Shot Video Question Answer MSRVTT-QA VideoChat2 Accuracy 54.1 #23 of 30 Archive leaderboard report
Zero-Shot Video Question Answer MSRVTT-QA VideoChat2 Confidence Score 3.3 #23 of 30 Archive leaderboard report
Zero-Shot Video Question Answer MSVD-QA VideoChat2 Accuracy 70.0 #18 of 28 Archive leaderboard report
Zero-Shot Video Question Answer MSVD-QA VideoChat2 Confidence Score 3.9 #18 of 28 Archive leaderboard report
Zero-Shot Video Question Answer NExT-QA VideoChat2 Accuracy 61.7 #20 of 27 Archive leaderboard report
Zero-Shot Video Question Answer STAR Benchmark VideoChat2 Accuracy 59.0 #1 of 4 Archive leaderboard report
Zero-Shot Video Question Answer TVQA VideoChat_HD_mistral (no speech) Accuracy 50.6 #4 of 9 Archive leaderboard report
Zero-Shot Video Question Answer TVQA VideoChat_mistral (no speech) Accuracy 46.4 #5 of 9 Archive leaderboard report
Zero-Shot Video Question Answer TVQA VideoChat2 (no speech) Accuracy 40.6 #6 of 9 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections