Datasets
M
OpenVals/MMLU-Pro
PreviewGate-readyBenchmark
Harder multi-choice knowledge benchmark across 14 domains.
Rows
12,032
Uses
980
Size bucket
10K-100K
Format
parquet
Modality
Text
Updated
Sep 22, 2026
Attach to a validation gate
Sign in to mount this dataset in a gate run, configure thresholds, and get an auditable pass/fail verdict.
More text datasets
- MOpenVals/MT-BenchMulti-turn conversation benchmark with 80 high-quality questions.2.3k
- JOpenVals/JailbreakBenchAdversarial jailbreak prompts covering DAN, obfuscation, and role-play attacks.2.0k
- TOpenVals/TruthfulQA817 questions testing whether models reproduce human falsehoods.1.2k
- OOpenVals/OWASP LLM Top-10Prompt sets mapped to every OWASP LLM risk category.1.1k