VRBench: A Benchmark for Multi-Step Reasoning in Long Narrative Videos

TL;DR

VRBench is a comprehensive benchmark for evaluating large models' multi-step reasoning in long narrative videos, addressing temporal reasoning and procedural validity with extensive datasets and a multi-phase evaluation pipeline.

Contribution

It introduces the first long narrative video benchmark with multi-step reasoning annotations and a novel evaluation framework for assessing reasoning chains in models.

Findings

12 LLMs evaluated with detailed analysis

19 VLMs assessed for multi-step reasoning capabilities

Proposed scoring metric evaluates reasoning quality comprehensively

Abstract

We present VRBench, the first long narrative video benchmark crafted for evaluating large models' multi-step reasoning capabilities, addressing limitations in existing evaluations that overlook temporal reasoning and procedural validity. It comprises 960 long videos (with an average duration of 1.6 hours), along with 8,243 human-labeled multi-step question-answering pairs and 25,106 reasoning steps with timestamps. These videos are curated via a multi-stage filtering process including expert inter-rater reviewing to prioritize plot coherence. We develop a human-AI collaborative framework that generates coherent reasoning chains, each requiring multiple temporally grounded steps, spanning seven types (e.g., event attribution, implicit inference). VRBench designs a multi-phase evaluation pipeline that assesses models at both the outcome and process levels. Apart from the MCQs for the final results, we propose a progress-level LLM-guided scoring metric to evaluate the quality of the reasoning chain from multiple dimensions comprehensively. Through extensive evaluations of 12 LLMs and 19 VLMs on VRBench, we undertake a thorough analysis and provide valuable insights that advance the field of multi-step reasoning.