NRT-Bench: leaderboard
Large language model (LLM) agents are increasingly proposed as supervisory components for safety-critical systems, yet their robustness under sustained, adaptive adversarial pressure remains poorly characterized. We present NRT-Bench, a benchmark for multi-turn red-teaming of LLM agents acting as operators of a safety-critical system, instantiated in a simulated nuclear power plant control room. A
Metric: Aggregate ASR (self-reported). Source: benchmarklist.com. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Haiku 4.5 | 8.7 | #280 |
| 2 | Qwen 3.5 35B A3B | 11.4 | #256 |
| 3 | GPT-5.4 Mini | 11.4 | #212 |
| 4 | Gemma 4 26B A4B (IT) | 12.1 | #297 |
Interactive version: theaggregate.ai/benchmark?slug=nrt-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.