SafeAtlas-Bench

What is SafeAtlas-Bench?

Evaluates multimodal safety moderation via 5,000-instance held-out set with five-level ordered risk predictions and continuous scores across 15 harm categories.

Released

2026-08-29

Evaluates

General AI, Multimodal, Safety, Safety & Trustworthiness

Openness

unknown

Importer

Claire Radar

Review status

ai-reviewed

Reported scores

0

Source provenance

• Original evidence https://arxiv.org/abs/2608.29098

SafeAtlas-Bench paper

• Paper https://arxiv.org/abs/2608.29098

Which sources cite SafeAtlas-Bench?

• SafeAtlas-Bench

Related General AI, Multimodal, Safety, Safety & Trustworthiness benchmarks

• $\tau^\tau$-Bench: An Environment for End-To-End, Realistic Agent Construction
• 3DCodeBench
• 5-Dialects-BN
• A-Bench
• A-OKVQA
• A3S-Bench
• AARRI-Bench
• ActBench