Can You Count to Nine? A Human Evaluation Benchmark for Counting Limits in Modern Text-to-Video Models

• URL: http://arxiv.org/abs/2504.04051v1
• Date: Sat, 05 Apr 2025 04:13:06 GMT
• Title: Can You Count to Nine? A Human Evaluation Benchmark for Counting Limits in Modern Text-to-Video Models
• Authors: Xuyang Guo, Zekai Huang, Jiayan Huo, Yingyu Liang, Zhenmei Shi, Zhao Song, Jiahao Zhang,
• Abstract summary: We present T2VCountBench, a specialized benchmark aiming at evaluating the counting capability of SOTA text-to-video models as of 2025.<n>Our experiments reveal that all existing models struggle with basic numerical tasks, almost always failing to generate videos with an object count of 9 or fewer.<n>Our findings highlight important challenges in current text-to-video generation and provide insights for future research aimed at improving adherence to basic numerical constraints.
• Score: 19.51519289698524
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Generative models have driven significant progress in a variety of AI tasks, including text-to-video generation, where models like Video LDM and Stable Video Diffusion can produce realistic, movie-level videos from textual instructions. Despite these advances, current text-to-video models still face fundamental challenges in reliably following human commands, particularly in adhering to simple numerical constraints. In this work, we present T2VCountBench, a specialized benchmark aiming at evaluating the counting capability of SOTA text-to-video models as of 2025. Our benchmark employs rigorous human evaluations to measure the number of generated objects and covers a diverse range of generators, covering both open-source and commercial models. Extensive experiments reveal that all existing models struggle with basic numerical tasks, almost always failing to generate videos with an object count of 9 or fewer. Furthermore, our comprehensive ablation studies explore how factors like video style, temporal dynamics, and multilingual inputs may influence counting performance. We also explore prompt refinement techniques and demonstrate that decomposing the task into smaller subtasks does not easily alleviate these limitations. Our findings highlight important challenges in current text-to-video generation and provide insights for future research aimed at improving adherence to basic numerical constraints.

Related papers

• Text-to-Image Diffusion Models Cannot Count, and Prompt Refinement Cannot Help [18.70937620674227]
  We introduce T2ICountBench, a novel benchmark designed to rigorously evaluate the counting ability of state-of-the-art text-to-image diffusion models. Our evaluations reveal that all state-of-the-art diffusion models fail to generate the correct number of objects, with accuracy dropping significantly as the number of objects increases.
  arXiv  Detail & Related papers  (2025-03-10T03:28:18Z)
• HumanVBench: Exploring Human-Centric Video Understanding Capabilities of MLLMs with Synthetic Benchmark Data [55.739633494946204]
  We present HumanVBench, an innovative benchmark meticulously crafted to bridge gaps in the evaluation of video MLLMs.<n>HumanVBench comprises 16 carefully designed tasks that explore two primary dimensions: inner emotion and outer manifestations, spanning static and dynamic, basic and complex, as well as single-modal and cross-modal aspects.<n>A comprehensive evaluation across 22 SOTA video MLLMs reveals notable limitations in current performance, especially in cross-modal and emotion perception.
  arXiv  Detail & Related papers  (2024-12-23T13:45:56Z)
• Rethinking Video-Text Understanding: Retrieval from Counterfactually Augmented Data [19.210471935816273]
  We propose a novel evaluation task for video-text understanding, namely retrieval from counterfactually augmented data (RCAD) and a new Feint6K dataset. To succeed on our new evaluation task, models must derive a comprehensive understanding of the video from cross-frame reasoning. Our approach successfully learn more discriminative action embeddings and improves results on Feint6K when applied to multiple video-text models.
  arXiv  Detail & Related papers  (2024-07-18T01:55:48Z)
• Chrono: A Simple Blueprint for Representing Time in MLLMs [34.036784478999245]
  We investigate the challenge of contextual and temporal comprehension in video-language models by exploring the task of temporal localization in videos.<n>We introduce Chrono, a universal sequence blueprint that can be applied to an image-text pretrained MLLM.<n>We achieve a new SOTA in moment retrieval on the most widely used benchmarks Charades-STA, QVHighlights, ActivityNet Captions, and grounded video question answering on NeXT-GQA.
  arXiv  Detail & Related papers  (2024-06-26T06:59:09Z)
• Towards Multi-Task Multi-Modal Models: A Video Generative Perspective [5.495245220300184]
  This thesis chronicles our endeavor to build multi-task models for generating videos and other modalities under diverse conditions. We unveil a novel approach to mapping bidirectionally between visual observation and interpretable lexical terms. Our scalable visual token representation proves beneficial across generation, compression, and understanding tasks.
  arXiv  Detail & Related papers  (2024-05-26T23:56:45Z)
• CinePile: A Long Video Question Answering Dataset and Benchmark [55.30860239555001]
  We present a novel dataset and benchmark, CinePile, specifically designed for authentic long-form video understanding. Our comprehensive dataset comprises 305,000 multiple-choice questions (MCQs), covering various visual and multimodal aspects. We fine-tuned open-source Video-LLMs on the training split and evaluated both open-source and proprietary video-centric LLMs on the test split of our dataset.
  arXiv  Detail & Related papers  (2024-05-14T17:59:02Z)
• Scaling Up Video Summarization Pretraining with Large Language Models [73.74662411006426]
  We introduce an automated and scalable pipeline for generating a large-scale video summarization dataset. We analyze the limitations of existing approaches and propose a new video summarization model that effectively addresses them. Our work also presents a new benchmark dataset that contains 1200 long videos each with high-quality summaries annotated by professionals.
  arXiv  Detail & Related papers  (2024-04-04T11:59:06Z)
• EvalCrafter: Benchmarking and Evaluating Large Video Generation Models [70.19437817951673]
  We argue that it is hard to judge the large conditional generative models from the simple metrics since these models are often trained on very large datasets with multi-aspect abilities. Our approach involves generating a diverse and comprehensive list of 700 prompts for text-to-video generation. Then, we evaluate the state-of-the-art video generative models on our carefully designed benchmark, in terms of visual qualities, content qualities, motion qualities, and text-video alignment with 17 well-selected objective metrics.
  arXiv  Detail & Related papers  (2023-10-17T17:50:46Z)
• Look, Remember and Reason: Grounded reasoning in videos with language models [5.3445140425713245]
  Multi-temporal language models (LM) have recently shown promising performance in high-level reasoning tasks on videos. We propose training an LM end-to-end on low-level surrogate tasks, including object detection, re-identification, tracking, to endow the model with the required low-level visual capabilities. We demonstrate the effectiveness of our framework on diverse visual reasoning tasks from the ACRE, CATER, Something-Else and STAR datasets.
  arXiv  Detail & Related papers  (2023-06-30T16:31:14Z)
• A Video Is Worth 4096 Tokens: Verbalize Videos To Understand Them In Zero Shot [67.00455874279383]
  We propose verbalizing long videos to generate descriptions in natural language, then performing video-understanding tasks on the generated story as opposed to the original video. Our method, despite being zero-shot, achieves significantly better results than supervised baselines for video understanding. To alleviate a lack of story understanding benchmarks, we publicly release the first dataset on a crucial task in computational social science on persuasion strategy identification.
  arXiv  Detail & Related papers  (2023-05-16T19:13:11Z)
• Look Before you Speak: Visually Contextualized Utterances [88.58909442073858]
  We create a task for predicting utterances in a video using both visual frames and transcribed speech as context. By exploiting the large number of instructional videos online, we train a model to solve this task at scale, without the need for manual annotations. Our model achieves state-of-the-art performance on a number of downstream VideoQA benchmarks.
  arXiv  Detail & Related papers  (2020-12-10T14:47:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.