VideoZeroBench: Probing the Limits of Video MLLMs with Spatio-Temporal Evidence Verification

2026-04-01

AI Summary

Current evaluations of video-based multimodal large language models often suffer from inflated scores and a lack of rigorous spatiotemporal evidence verification, making it difficult to assess fine-grained reasoning capabilities. To address this, this work proposes VideoZeroBench—a hierarchical benchmark for long-form video question answering comprising 500 human-annotated questions, each paired with precise temporal intervals and spatial bounding boxes as verifiable evidence. The benchmark introduces a five-level progressive evaluation protocol that incrementally constrains answer generation, temporal localization, and spatial grounding in a joint manner. For the first time, it incorporates a strict spatiotemporal evidence alignment mechanism, revealing a critical gap in current models’ capacity for verifiable reasoning: while standard QA accuracy remains below 17%, the accuracy drops to under 1% across all models when requiring both correct answers and precise spatiotemporal localization.

Abstract

Recent video multimodal large language models achieve impressive results across various benchmarks. However, current evaluations suffer from two critical limitations: (1) inflated scores can mask deficiencies in fine-grained visual understanding and reasoning, and (2) answer correctness is often measured without verifying whether models identify the precise spatio-temporal evidence supporting their predictions. To address this, we present VideoZeroBench, a hierarchical benchmark designed for challenging long-video question answering that rigorously verifies spatio-temporal evidence. It comprises 500 manually annotated questions across 13 domains, paired with temporal intervals and spatial bounding boxes as evidence. To disentangle answering generation, temporal grounding, and spatial grounding, we introduce a five-level evaluation protocol that progressively tightens evidence requirements. Experiments show that even Gemini-3-Pro correctly answers fewer than 17% of questions under the standard end-to-end QA setting (Level-3). When grounding constraints are imposed, performance drops sharply: No model exceeds 1% accuracy when both correct answering and accurate spatio-temporal localization are required (Level-5), with most failing to achieve any correct grounded predictions. These results expose a significant gap between surface-level answer correctness and genuine evidence-based reasoning, revealing that grounded video understanding remains a bottleneck for long-video QA. We further analyze performance across minimal evidence spans, atomic abilities, and inference paradigms, providing insights for future research in grounded video reasoning. The benchmark and code will be made publicly available.

Problem

• video MLLMs
• spatio-temporal evidence
• grounded reasoning
• long-video QA
• evaluation benchmark

Innovation

• spatio-temporal grounding
• video multimodal LLMs
• evidence verification
• hierarchical benchmark
• long-video QA