Share your thoughts, 1 month free Claude Pro on usSee more

Question Answering on Natural Questions (Accuracy)

46.39Accuracy

RL

Updated 23d ago

Evaluation Results

Method	Links
RL 2026.05		46.39
InternLM2-20B 2024.03		44.6
InternLM2-7B 2024.03		42.2
Mixtral-8x7B-v0.1 2024.03		39.6
ChatGLM3-6B-Base 2024.03		38.7
Qwen-14B 2024.03		37.1
RL 2026.05		36.91
RFT 2026.05		36.7
Llama2-13B 2024.03		34.5
RL 2026.05		34.12
LLAMA2 2024.04		31.2
Mistral-7B-v0.1 2024.03		31.1
DPO 2026.05		30.48
InternLM2-20B-Base 2024.03		30.3
Base 2026.05		30.15
SFT 2026.05		29.45
Llama2-7B 2024.03		28.8
InternLM2-7B-Base 2024.03		28.4
Baichuan2-13B-Base 2024.03		27.5
RFT 2026.05		27.03
LLAMA2 2024.04		25.7
MEGALODON 2024.04		25.7
SFT 2026.05		25.39
RFT 2026.05		25.33
Ours (SCD) 2025.02		24.3
Baichuan2-7B-Base 2024.03		24
SFT 2026.05		23.67
Mistral 2024.04		23.2
DPO 2026.05		23.06
Gemma 2024.04		23
DPO 2026.05		22.94
Base 2026.05		22.79
Qwen-7B 2024.03		22.7
Greedy 2025.02		22.1
Base 2026.05		21.42
MPT 2024.04		20.8