Scientific Reasoning on GPQA (Accuracy)

83.8Accuracy

Nanbeige4.1-3B

Updated 1mo ago

Evaluation Results

Method	Links
Nanbeige4.1-3B 2026.02		83.8
Nanbeige4-3B-2511 2026.02		82.2
Qwen3-Next-80B-A3B 2026.02		77.2
Qwen3-30B-A3B-2507 2026.02		73.4
Qwen3-32B 2026.02		68.4
Qwen3-4B-2507 2026.02		65.8
Poisoned-RM (Thought-Transfer) 2026.01		51
Poisoned-RM (Thought-Transfer) 2026.01		45.5
Qwen-14B (Base Model) 2026.01		39.4
Qwen-14B (Base Model) 2026.01		39.4
Ouro2.6B-Thinking + RLTT 2026.02		38.4
BF16 2026.02		38.38
GPTQ 2026.02		36.86
OJBKQ 2026.02		36.36
BF16 2026.02		35.98
OJBKQ 2026.02		34.34
AWQ 2026.02		33.33
BF16 2026.02		33.33
OJBKQ 2026.02		32.93
AWQ 2026.02		32.83
AWQ 2026.02		32.32
GPTQ 2026.02		31.81
QUIP 2026.02		31.31
EB-Sampler 2025.12		29.9
QUIP 2026.02		29.87
GPTQ 2026.02		29.27
Uniform 2025.12		29
Confidence 2025.12		29
PC-Sampler w/ E-SMC 2025.12		29
PC-Sampler w/ E-SMC 2025.12		28.9
PC-Sampler w/ E-BON 2025.12		28.9
PC-Sampler w/ E-BON 2025.12		28.8
Entropy 2025.12		28.8
P2 2025.12		28.8
PC-Sampler 2025.12		28.8
Fast-dLLM 2025.12		28.6
PC-Sampler 2025.12		28.6
Margin 2025.12		28.6
EB-Sampler 2025.12		28.6
Entropy 2025.12		28.4
Margin 2025.12		28.4
P2 2025.12		28.2
Uniform 2025.12		28.1
Confidence 2025.12		27.9
Fast-dLLM 2025.12		27.9
Semi-AR 2025.12		27.7
Semi-AR 2025.12		26.1
QUIP 2026.02		25.75
Ouro2.6B-Thinking + GRPO 2026.02		19.7
Ouro2.6B-Thinking 2026.02		18.7
Ouro2.6B-Thinking + SFT 2026.02		17.7
Qwen3 – 4B 2026.02		11.1
DeepSeekR1 – 7B 2026.02		11.1
DeepSeekR1 – 1.5B 2026.02		6.57
Qwen3 – 1.7B 2026.02		4.55