Benchmark Radar
AI BENCHMARK PROFILE

FIRM-Video-Bench

General AIMultimodal Perception

FIRM-Video-Bench is a benchmark of 250 text-to-video generations with 750 point-wise human annotations covering Instruction Following, World Coherence, and Perceptual Quality.

Released
2026-08-22
Readiness
Paper only
Primary field
General AI

Why it matters

It provides a fixed annotated set for evaluating reward models' ability to score videos on fine-grained dimensions, enabling comparison of text-to-video evaluation approaches.

Motivation

Reliable reward models are essential for text-to-video evaluation and alignment.

Primary resources

Benchmark Radar records only publicly supported details and links back to primary sources for verification.