FIOVA: A Multi-Annotator Benchmark for Human-Aligned Video Captioning

TL;DR

FIOVA introduces a multi-annotator, human-centric benchmark and a new event-level evaluation metric for assessing the alignment of video captioning models with human perception, addressing limitations of existing benchmarks.

Contribution

The paper presents FIOVA, a novel multi-annotator benchmark with a new evaluation metric, enabling detailed analysis of model alignment with human understanding in video captioning.

Findings

FIOVA captures semantic diversity and agreement among annotators.

Evaluation reveals gaps in model performance across complexity levels.

Structural issues like event under-description are identified.

Abstract

Despite rapid progress in large vision-language models (LVLMs), existing video caption benchmarks remain limited in evaluating their alignment with human understanding. Most rely on a single annotation per video and lexical similarity-based metrics, failing to capture the variability in human perception and the cognitive importance of events. These limitations hinder accurate diagnosis of model capabilities in producing coherent, complete, and human-aligned descriptions. To address this, we introduce FIOVA (Five-In-One Video Annotations), a human-centric benchmark tailored for evaluation. It comprises 3,002 real-world videos (about 33.6s each), each annotated independently by five annotators. This design enables modeling of semantic diversity and inter-subjective agreement, offering a richer foundation for measuring human-machine alignment. We further propose FIOVA-DQ, an event-level evaluation metric that incorporates cognitive weights derived from annotator consensus, providing fine-grained assessment of event relevance and semantic coverage. Leveraging FIOVA, we conduct a comprehensive evaluation of nine representative LVLMs and introduce a complexity-aware analysis framework based on inter-annotator variation (CV). This reveals consistency gaps across difficulty levels and identifies structural issues such as event under-description and template convergence. Our results highlight FIOVA's diagnostic value for understanding LVLM behavior under varying complexity, setting a new standard for cognitively aligned evaluation in long-video captioning. The benchmark, annotations, metric, and model outputs are publicly released to support future evaluation-driven research in video understanding. More detailed information can be found at https://huuuuusy.github.io/fiova/.