OpenResearchAlpha

Artificial Intelligence for Security and Defence Applications IV · 12 Nov 2026

Evaluating vision-language models against automatic metrics for video captioning

Navya Rachakonda, George Awad

0
Citations
no database records one yet
0
Citations in 2026
so far
–
Leading author's h-index
no author has a profile yet
0
References
works it cites, from open citation data

Citations by database

Each database indexes a different set of journals, preprints and books, so they are shown side by side rather than blended.

  • OpenCitations0
  • Crossref0

Citations per year

The current year is still counting.

  • 2026 so far0

Abstract

The task of captioning videos has several applications, such as in defense and security video analysis, human-robot interaction, a visual tool for the hearing/visually impaired, video summarization, automatic scene description, and supporting video indexing.<sup>1, 2</sup> Traditional video caption evaluation has been adopting text-based metrics such as BLEU, METEOR, and CIDER.<sup>3</sup> However, one key limiting feature is that they often fail to capture the semantic alignment with the visual content, that is, their output does not always align with the image content. Consequently, this limitation was addressed by exploring whether vision-language models (VLMs), in particular CLIP, ALIGN, and LLAMA 3.2 11B Vision(LLAMA),<sup>4</sup> can serve as alternative evaluation tools. In this work, we propose a reference-free evaluation approach that extracts keyframes from videos, computes similarity scores between keyframes and captions using VLMs, and analyzes their correlation with traditional metrics. We used a benchmark dataset of 1,757 videos and 31,626 captioning systems’ outputs. We observe weak-to-moderate correlations between VLM-based similarity scores and traditional evaluation metrics. Our findings suggest that while VLMs can capture aspects of semantic alignment that traditional metrics cannot capture, they are still not able to fully replace conventional approaches just yet.

What the data shows

  • No database records a citation yet, as is usual for a paper this new.

Graph

This paper in the middle, its authors and their most-cited papers below, the papers it cites to the left and the papers citing it to the right. The timeline lays every paper out by year.

Citing paper shares an authorAuthor without a profileSize shows citations, or h-index for authors

Select a node for its details. Drag to pan, scroll to zoom, drag a node to move it.

Start with

Authors (2)

Each author's standing in QAI's scholar index: the works their ORCID iD is attached to, and for anyone whose ORCID record names a lab OpenResearch follows, their arXiv preprints matched by name and a shared co-author. Names are as printed on the paper.

ProfileMost-cited paper

Cited by 0 papers

Who cites this paper, from open citation data, with Crossref's record of each; the 100 newest when there are more.

0 shown
No citing papers for this filter.

References (0)

What it cites, from open citation data, with Crossref's record of each.

No references in the open citation data yet.
Evaluating vision-language models against automatic metrics for video captioning - QAI OpenResearch