List your company
ResearchFormerly WizWand

Safety Evaluation on Trotter

30.8OverRefusal ScoreBase
10.31215.63120.9526.269May 9, 2026
Updated 5mo ago

Evaluation results

MethodLinks
BaseModel Backbone=DeepSeek-R1-Distill-Qwen-14B, RL Training (GRPO)=true2026.0530.8
STAR-1Model Backbone=DeepSeek-R1-Distill-Qwen-14B, RL Training (GRPO)=true2026.0528.8
SafeChainModel Backbone=DeepSeek-R1-Distill-Qwen-7B, RL Training (GRPO)=true2026.0527.3
STAR-1Model Backbone=DeepSeek-R1-Distill-Qwen-7B, RL Training (GRPO)=true2026.0526.3
BaseModel Backbone=DeepSeek-R1-Distill-Qwen-7B, RL Training (GRPO)=true2026.0523.7
SInternalModel Backbone=DeepSeek-R1-Distill-Qwen-7B, RL Training (GRPO)=true2026.0523.7
SafeChainModel Backbone=DeepSeek-R1-Distill-Qwen-14B, RL Training (GRPO)=true2026.0523.7
STAR-1Model Backbone=DeepSeek-R1-Distill-Llama-8B, RL Training (GRPO)=true2026.0521.7
SInternalModel Backbone=DeepSeek-R1-Distill-Qwen-14B, RL Training (GRPO)=true2026.0521.7
SafeChainModel Backbone=DeepSeek-R1-Distill-Llama-8B, RL Training (GRPO)=true2026.0518.7
BaseModel Backbone=DeepSeek-R1-Distill-Llama-8B, RL Training (GRPO)=true2026.0514.4
SInternalModel Backbone=DeepSeek-R1-Distill-Llama-8B, RL Training (GRPO)=true2026.0511.1