titleFVScore: Specification-Driven Neuro-Symbolic Video Evaluation
authorsDo Xuan Long, Long T. Le, Yiwen Song, Min-Yen Kan, Tomas Pfister, Yale Song
date2026-11-17
venuearXiv (preprint, not peer reviewed)
idarXiv:2611.22673
affiliationGoogle
upvotes4 (Hugging Face Daily Papers, 2026-11-21)
linksabs pdf hf
cite
relatedRRSI: Regularized Recursive Self-Improvement of Agent Harnesses
Dream-RSI: Recursive Self-Improvement through Evolving Worlds
EnvHarness: Awakening Static Worlds for Agent Learning
StreamVibe: Proactive Video Agents for Evolving User Intent
Harness^2: Recursive Agent Harnessing for an Open World
abstractWe introduce Faithfulness Video Score (FVSCORE), a neuro-symbolic framework for specification-driven video generation evaluation. Existing evaluators rely on holistic MLLM judgments or vision-language metrics, yet aligning their scores with human judgments and tracing them to specific temporal violations remain open challenges. FVSCORE addresses these limitations by compiling each video into explicit propositions capturing prompt requirements and implicit propositions capturing unstated physical, temporal, and cross-modal constraints. It grounds each proposition using specialized perceptual tools and evaluates the resulting confidence traces with soft temporal operators inspired by Linear Temporal Logic. A Significance-Aware Calibration step then aligns verification with human perception by retaining only perceptually significant violations before producing a timestamp- and evidence-grounded report. We conduct extensive experiments across public and complex multi-event benchmarks with the strongest video models to date, demonstrating that FVSCORE outperforms holistic and agent-based evaluators in human correlation while providing highly accurate and interpretable feedback.

bibtex

@misc{long2026fvscore,
  title={FVScore: Specification-Driven Neuro-Symbolic Video Evaluation},
  author={Do Xuan Long and Long T. Le and Yiwen Song and Min-Yen Kan and Tomas Pfister and Yale Song},
  year={2026},
  eprint={2611.22673},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2611.22673}
}