Neuro-Symbolic Evaluation of Text-to-Video Models using Formal Verification

TL;DR

This paper introduces NeuS-V, a formal verification-based metric for evaluating text-to-video models, focusing on temporal fidelity and alignment, and demonstrates its superior correlation with human judgment over existing metrics.

Contribution

The paper presents NeuS-V, a novel neuro-symbolic evaluation method that rigorously assesses text-to-video alignment using formal verification techniques, addressing limitations of current metrics.

Findings

NeuS-V correlates over 5x better with human evaluations than existing metrics.

Current models perform poorly on temporally complex prompts.

A new dataset of temporally extended prompts is introduced for benchmarking.

Abstract

Recent advancements in text-to-video models such as Sora, Gen-3, MovieGen, and CogVideoX are pushing the boundaries of synthetic video generation, with adoption seen in fields like robotics, autonomous driving, and entertainment. As these models become prevalent, various metrics and benchmarks have emerged to evaluate the quality of the generated videos. However, these metrics emphasize visual quality and smoothness, neglecting temporal fidelity and text-to-video alignment, which are crucial for safety-critical applications. To address this gap, we introduce NeuS-V, a novel synthetic video evaluation metric that rigorously assesses text-to-video alignment using neuro-symbolic formal verification techniques. Our approach first converts the prompt into a formally defined Temporal Logic (TL) specification and translates the generated video into an automaton representation. Then, it evaluates the text-to-video alignment by formally checking the video automaton against the TL specification. Furthermore, we present a dataset of temporally extended prompts to evaluate state-of-the-art video generation models against our benchmark. We find that NeuS-V demonstrates a higher correlation by over 5x with human evaluations when compared to existing metrics. Our evaluation further reveals that current video generation models perform poorly on these temporally complex prompts, highlighting the need for future work in improving text-to-video generation capabilities.