Benchmark RadarAI BENCHMARK PROFILE
SafeGen-Bench
Evaluates safety of conditional text-to-video generation using selected start frames and text prompts across 10 malicious categories, measuring unsafety scores and guardrail effectiveness.
- Released
- 2026-05-31
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Addresses the gap of safety evaluation when both text and image inputs are benign but output is harmful. Provides a benchmark to improve model safeguards in dynamic video generation.
Motivation
With the rapid advancements in text-to-image diffusion models, generative video models (T2V models) like Sora can now produce short synthetic videos from a text prompt or an initial image.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.