List your company
ResearchFormerly WizWand

Information Retrieval on Korean Web Search Offline Evaluation set (test)

38.56P@3AuthGR (Full)
6.99615.190523.38531.5795Apr 15, 2026
Updated 6mo ago

Evaluation results

MethodLinks
AuthGR (Full)Strategy=CPT + SFT + GRPO, Size=3B2026.0438.5654.6471.75
HyperCLOVAXStrategy=Supervised Fine-tuning (SFT), Size=14B2026.0438.5455.0872.89
AuthGR (CPT+SFT)Strategy=Continued Pre-training (CPT) + Supervised Fine-tuning (SFT), Size=3B2026.0437.2552.9370.31
LLaMA 3.2Strategy=Supervised Fine-tuning (SFT), Size=3B2026.0436.0251.0868.92
Qwen3Strategy=Supervised Fine-tuning (SFT), Size=4B2026.0435.8150.5368.43
HyperCLOVAXStrategy=Supervised Fine-tuning (SFT), Size=1.5B2026.0435.7350.5867.08
AuthGR (SFT)Strategy=Supervised Fine-tuning (SFT), Size=3B2026.0435.5550.5868.99
LLaMA 3.2Strategy=Supervised Fine-tuning (SFT), Size=1B2026.0434.7949.4866.79
HyperCLOVAXStrategy=Supervised Fine-tuning (SFT), Size=0.5B2026.0434.749.3366.34
Qwen3Strategy=Supervised Fine-tuning (SFT), Size=1.7B2026.0434.3348.5365.82
T5Gemma 2Strategy=Supervised Fine-tuning (SFT), Size=2B2026.0433.9948.0563.84
T5Gemma 2Strategy=Supervised Fine-tuning (SFT), Size=0.5B2026.0432.846.4661.51
GPT-4oStrategy=In-context Learning (ICL), Size=-2026.041723.4831.7
LLaMA 3.1Strategy=In-context Learning (ICL), Size=405B2026.0414.1319.7425.9
DeepSeek-V3.2Strategy=In-context Learning (ICL), Size=685B2026.0413.9820.2727.29
K-EXAONEStrategy=In-context Learning (ICL), Size=236B2026.0413.6619.1826.56
DeepSeek-V3Strategy=In-context Learning (ICL), Size=671B2026.0413.5919.3226.26
LLaMA 4 MaverickStrategy=In-context Learning (ICL), Size=400B2026.0412.7418.4124.83
Gemma 3Strategy=In-context Learning (ICL), Size=27B2026.0412.5517.3222.85
LLaMA 4 ScoutStrategy=In-context Learning (ICL), Size=109B2026.0410.6615.5521.28
DeepSeek-R1Strategy=In-context Learning (ICL), Size=671B2026.048.9113.2817.18
EXAONE 3.5Strategy=In-context Learning (ICL), Size=32B2026.048.2512.5316.12
Qwen3Strategy=In-context Learning (ICL), Size=32B2026.048.2111.7615.7