List your company

Safety Evaluation on XSTest (ASR)

0.9ASRAMALIA-9B-DPO
06.07512.1518.225Mar 27, 2026

Evaluation results

MethodLinks
AMALIA-9B-DPOModel category=Fully open models, Instruction-tuned=true, Post-training method=DPO, Judge model=Qwen3Guard-8B2026.030.9
Gemma 2-9BModel category=Open weight models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.031.4
EuroLLM-9BModel category=Fully open models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.031.9
AMALIA-9B-SFTModel category=Fully open models, Instruction-tuned=true, Post-training method=SFT, Judge model=Qwen3Guard-8B2026.031.9
Qwen 2.5-7BModel category=Open weight models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.032.4
Llama 3.1-8BModel category=Open weight models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.032.9
Gemma 3-12BModel category=Open weight models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.033.4
Apertus-8BModel category=Fully open models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.033.9
Gervasio-8BModel category=Open weight models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.033.9
Qwen 3-8BModel category=Open weight models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.034.4
Ministral-8BModel category=Open weight models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.039.9
Mistral-7BModel category=Open weight models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.0313.4
OLMo 2-7BModel category=Fully open models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.0317.4
Salamandra-7BModel category=Fully open models, Instruction-tuned=true, Judge model=Qwen3Guard-8B2026.0323.4