List your company
ResearchFormerly WizWand

Open-ended Generation on AlpacaEval

93.5Win Rate vs Davinci-003GPT-4
1.4625.35549.2573.145Jun 7, 2023Nov 24, 2023May 12, 2024Oct 29, 2024Apr 17, 2025Oct 4, 2025Mar 23, 2026
Updated 7mo ago

Evaluation results

MethodLinks
GPT-42023.0693.5
ChatGPT2023.0683.6
RED# Param=0.26M2024.0281.69
LoRA# Param=8.39M2024.0281.48
FT# Param=6739M2024.0280.93
ShareGPT 65B2023.0673.6
TÜLU-1.1 13B2023.0664
TÜLU 30B2023.0662.3
TÜLU 65B2023.0661.8
TÜLU-1.1 7B2023.0657.3
TÜLU 13B2023.0656.5
TÜLU 7B2023.0648.6
Human mix. 65B2023.0643.4
DPOBackbone=Qwen3-8B2026.0324.6
Prompt EngBackbone=Gemma-2-9B2026.0322.51
InternLM2-Chat-20BCategory=~20B Models, Parameter Size=20B, Strategy=RLHF2024.0321.8
Base ModelBackbone=Gemma-2-9B2026.0321.12
DSPABackbone=Gemma-2-9B2026.0320
DSPABackbone=Gemma-2-2B2026.0319.3
RepEBackbone=Gemma-2-9B2026.0318.88
DSPABackbone=Qwen3-8B2026.0317.83
Base ModelBackbone=Gemma-2-2B2026.0317.54
DPOBackbone=Gemma-2-9B2026.0317.39
RepEBackbone=Gemma-2-2B2026.0316.94
DPOBackbone=Gemma-2-2B2026.0315.82
Prompt EngBackbone=Gemma-2-2B2026.0315.28
Mistral-7B-Instruct-v0.2Category=~7B Models, Parameter Size=7B2024.0314.7
Static-SAEBackbone=Gemma-2-9B2026.0313.56
Base ModelBackbone=Qwen3-8B2026.0313.47
RepEBackbone=Qwen3-8B2026.0312.95
Prompt EngBackbone=Qwen3-8B2026.0311.74
InternLM2-Chat-7BCategory=~7B Models, Parameter Size=7B, Strategy=RLHF2024.0311.3
GPT-3.5Category=API Models2024.039.6
Static-SAEBackbone=Gemma-2-2B2026.039.33
Static-SAEBackbone=Qwen3-8B2026.038.58
InternLM2-Chat-20B-SFTCategory=~20B Models, Parameter Size=20B, Strategy=SFT2024.038
Llama2-13B-ChatCategory=~20B Models, Parameter Size=13B2024.037.7
Qwen-14B-ChatCategory=~20B Models, Parameter Size=14B2024.037.5
InternLM2-Chat-7B-SFTCategory=~7B Models, Parameter Size=7B, Strategy=SFT2024.036.9
Llama2-7B-ChatCategory=~7B Models, Parameter Size=7B2024.035