Sora Detector: A Unified Hallucination Detection for Large Text-to-Video Models
Zhixuan Chu, Lei Zhang, Yichen Sun, Siqiao Xue, Zhibo Wang, Zhan Qin, Kui Ren
TL;DR
This work tackles hallucinations in large text-to-video models by introducing SoraDetector, a unified detection framework that jointly analyzes prompt-consistency, static frame-level anomalies, and dynamic temporal inconsistencies. The pipeline combines keyframe-based content summarization, static and dynamic knowledge graphs, and multimodal large language models to detect three hallucination types and generate a comprehensive video quality report via the Sora Detector Agent. A novel benchmark, T2VHaluBench, is proposed to enable rigorous evaluation of T2V hallucination detectors across diverse models and prompts. Experimental results show strong overall precision (>98%) and reveal areas for improvement in static and dynamic categories, with ablations demonstrating the value of knowledge graphs for improved detection. The methodology offers a practical, scalable tool for ensuring reliability in video generation and provides a benchmark resource to drive future advancements in T2V hallucination detection.
Abstract
The rapid advancement in text-to-video (T2V) generative models has enabled the synthesis of high-fidelity video content guided by textual descriptions. Despite this significant progress, these models are often susceptible to hallucination, generating contents that contradict the input text, which poses a challenge to their reliability and practical deployment. To address this critical issue, we introduce the SoraDetector, a novel unified framework designed to detect hallucinations across diverse large T2V models, including the cutting-edge Sora model. Our framework is built upon a comprehensive analysis of hallucination phenomena, categorizing them based on their manifestation in the video content. Leveraging the state-of-the-art keyframe extraction techniques and multimodal large language models, SoraDetector first evaluates the consistency between extracted video content summary and textual prompts, then constructs static and dynamic knowledge graphs (KGs) from frames to detect hallucination both in single frames and across frames. Sora Detector provides a robust and quantifiable measure of consistency, static and dynamic hallucination. In addition, we have developed the Sora Detector Agent to automate the hallucination detection process and generate a complete video quality report for each input video. Lastly, we present a novel meta-evaluation benchmark, T2VHaluBench, meticulously crafted to facilitate the evaluation of advancements in T2V hallucination detection. Through extensive experiments on videos generated by Sora and other large T2V models, we demonstrate the efficacy of our approach in accurately detecting hallucinations. The code and dataset can be accessed via GitHub.