List your company
ResearchFormerly WizWand

Real-world Query Evaluation on WildBench

71.5WildBench Accuracy+RL (Skywork-Reward-V2-Llama-3.1-8B)
51.01256.33161.6566.969Jul 2, 2025
Updated 8mo ago

Evaluation results

MethodLinks
+RL (Skywork-Reward-V2-Llama-3.1-8B)Model=Qwen2.5-7B2025.0771.5
Instruct (official)Model=Qwen2.5-7B2025.0770.9
+RL (Skywork-Reward-V2-Llama-3.1-8B)Model=Llama-3.1-8B2025.0770.2
+RL (Skywork-Reward-V2-Qwen3-4B)Model=Qwen2.5-7B2025.0769
+RL (Skywork-Reward-Gemma-2-27B-v0.2)Model=Qwen2.5-7B2025.0767.8
+RL (Skywork-Reward-V2-Qwen3-4B)Model=Llama-3.1-8B2025.0765
+RL (Skywork-Reward-Llama-3-8B-v0.2)Model=Qwen2.5-7B2025.0764.9
Instruct (official)Model=Llama-3.1-8B2025.0764.2
+RL (Skywork-Reward-Gemma-2-27B-v0.2)Model=Llama-3.1-8B2025.0761.5
+SFTModel=Qwen2.5-7B2025.0760.5
+SFTModel=Llama-3.1-8B2025.0760.3
+RL (Skywork-Reward-Llama-3-8B-v0.2)Model=Llama-3.1-8B2025.0757.8
BaseModel=Llama-3.1-8B2025.0754.9
BaseModel=Qwen2.5-7B2025.0751.8