SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2505.14640 · 2025

VideoEval-Pro: Robust and Realistic Long Video Understanding Evaluation

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 14 functions out of this paper's own repositories and ran 7 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
opengvlab/videochat-flash canonical 5 of 7
opengvlab/internvideo2 canonical 2 of 7
FunctionStatusWhere it lives
repeat_kv Ran opengvlab/videochat-flash/llava-train_videochat/llava/model/language_model/modeling_qwen2_flash.py
code served (permissive licence) · get_code("30d7eec482ebf6b1")
apply_rotary_pos_emb Ran opengvlab/videochat-flash/llava-train_videochat/llava/model/language_model/modeling_qwen2_flash.py
code served (permissive licence) · get_code("d61c483a3c2b3156")
extract_patches Ran opengvlab/videochat-flash/llava-train_videochat/llava/mm_utils.py
code served (permissive licence) · get_code("b0a851cae92754ff")
find_all_linear_names Ran opengvlab/internvideo2/xtuner-train_internvideo2_5/xtuner/model/utils.py
code served (permissive licence) · get_code("53281c424cf5891f")
packing_collate Ran opengvlab/internvideo2/xtuner-train_internvideo2_5/unify_internvl2_train_r16.py
code served (permissive licence) · get_code("e6bb171e5f95a784")
rotate_half Ran opengvlab/videochat-flash/llava-train_videochat/llava/model/language_model/modeling_qwen2_flash.py
code served (permissive licence) · get_code("b99eea6376d1e212")
unpad_image Ran opengvlab/videochat-flash/llava-train_videochat/llava/model/llava_arch.py
code served (permissive licence) · get_code("2c7b71210b8a363f")
auto_pad_images Not yet run opengvlab/videochat-flash/llava-train_videochat/llava/mm_utils.py
code served (permissive licence) · get_code("3c8823948724e269")
build_model Not yet run opengvlab/internvideo2/xtuner-train_internvideo2_5/xtuner/apis/model.py
code served (permissive licence) · get_code("dcdc7ef233f3e1e8")
convert_state_dict_to_hf Not yet run opengvlab/internvideo2/xtuner-train_internvideo2_5/xtuner/model/llava.py
code served (permissive licence) · get_code("f8035a195671164a")
create_reference_model Not yet run opengvlab/internvideo2/xtuner-train_internvideo2_5/xtuner/model/dpo.py
code served (permissive licence) · get_code("269c10dd16c744e8")
get_peft_model_state_dict Not yet run opengvlab/internvideo2/xtuner-train_internvideo2_5/xtuner/model/utils.py
code served (permissive licence) · get_code("9356d2ed9b73134b")
reduce_mean Not yet run opengvlab/internvideo2/xtuner-train_internvideo2_5/xtuner/model/reward.py
code served (permissive licence) · get_code("a0a407c2a1c1f7d3")
resize_and_center_crop Not yet run opengvlab/videochat-flash/llava-train_videochat/llava/mm_utils.py
code served (permissive licence) · get_code("c80358dff96c25ce")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Large multimodal models (LMMs) have recently emerged as a powerful tool for long video understanding (LVU), prompting the development of standardized LVU benchmarks to evaluate their performance. However, our investigation reveals a rather sober lesson for existing LVU benchmarks. First, most existing benchmarks rely heavily on multiple-choice questions (MCQs), whose evaluation results are inflated due to the possibility of guessing the correct answer; Second, a significant portion of questions in these benchmarks have strong priors to allow models to answer directly without even reading the input video. For example, Gemini-1.5-Pro can achieve over 50\% accuracy given a random frame from a long video on Video-MME. We also observe that increasing the number of frames does not necessarily lead to improvement on existing benchmarks, which is counterintuitive. As a result, the validity and robustness of current LVU benchmarks are undermined, impeding a faithful assessment of LMMs' long-video understanding capability. To tackle this problem, we propose VideoEval-Pro, a realistic LVU benchmark containing questions with open-ended short-answer, which truly require understanding the entire video. VideoEval-Pro assesses both segment-level and full-video understanding through perception and reasoning tasks. By evaluating 21 proprietary and open-source video LMMs, we conclude the following findings: (1) video LMMs show drastic performance ($>$25\%) drops on open-ended questions compared with MCQs; (2) surprisingly, higher MCQ scores do not lead to higher open-ended scores on VideoEval-Pro; (3) compared to other MCQ benchmarks, VideoEval-Pro benefits more from increasing the number of input frames. Our results show that VideoEval-Pro offers a more realistic and reliable measure of long video understanding, providing a clearer view of progress in this domain.

For agents

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

get_harvested_code_for_paper("2505.14640")
get_code_for_paper("2505.14640")
have("2505.14640")

Connect an agent — have() is free.