MME-CoF-Pro: Evaluating Reasoning Coherence in Video Generative Models with Text and Visual Hints

TL;DR

This paper introduces MME-CoF-Pro, a benchmark to evaluate reasoning coherence in video generative models, revealing their weak reasoning abilities and the effects of different hints on their performance.

Contribution

It presents a new comprehensive benchmark with evaluation metrics and settings to assess reasoning coherence in video models, addressing a gap in current evaluation methods.

Findings

Video models show weak reasoning coherence, independent of generation quality.

Text hints improve correctness but can cause inconsistencies and hallucinations.

Visual hints help with perceptual tasks but are less effective for fine-grained reasoning.

Abstract

Video generative models show emerging reasoning behaviors. It is essential to ensure that generated events remain causally consistent across frames for reliable deployment, a property we define as reasoning coherence. To bridge the gap in literature for missing reasoning coherence evaluation, we propose MME-CoF-Pro, a comprehensive video reasoning benchmark to assess reasoning coherence in video models. Specifically, MME-CoF-Pro contains 303 samples across 16 categories, ranging from visual logical to scientific reasoning. It introduces Reasoning Score as evaluation metric for assessing process-level necessary intermediate reasoning steps, and includes three evaluation settings, (a) no hint (b) text hint and (c) visual hint, enabling a controlled investigation into the underlying mechanisms of reasoning hint guidance. Evaluation results in 7 open and closed-source video models reveals insights including: (1) Video generative models exhibit weak reasoning coherence, decoupled from generation quality. (2) Text hints boost apparent correctness but often cause inconsistency and hallucinated reasoning (3) Visual hints benefit structured perceptual tasks but struggle with fine-grained perception. Website: https://video-reasoning-coherence.github.io/