Knowledge Evaluation on GPQA

90.4Accuracy

CoT2-Meta

Updated 18d ago

Evaluation Results

Method	Links
CoT2-Meta 2026.03		90.4
ReST-MCTS* 2026.03		85.2
Vanilla ToT 2026.03		83.5
Best-of-16 2026.03		79.6
Greedy CoT 2026.03		74.2
DeepSeek-R1-Distill-Qwen-32B (Reasoning) 2026.01		59.39
Llama-3.3-70B-Instruct 2026.01		56.5
ReasonAny 2026.01		56.25
DictaLM 3.0 24B-Think 2026.02		55.13
Phi-4 2026.01		50.9
Reasoning 2026.01		47.51
LED 2026.01		46.97
Gemma 3 27B 2026.02		45
Mistral Small 3.1 2026.02		44.87
Task Arithmetic 2026.01		42.42
Qwen2.5-32B-Instruct (Safety) 2026.01		41.67
ReasonAny 2026.01		41.06
Linear 2026.01		40.1
LED 2026.01		39.38
Linear 2026.01		38.64
DARE 2026.01		38.64
Task Arithmetic 2026.01		36.36
DARE 2026.01		36.36
Qwen3-4B 2025.11		34.85
Multi Model 2026.02		34.3
TIES 2026.01		34.09
ICaRus 2026.02		33.8
FuseLLM 2026.01		33.33
TIES 2026.01		33.33
Foundation-Sec-8B-Instruct 2026.01		31.9
Foundation-Sec-8B-Reasoning 2026.01		31.7
Llama-3.2-3B 2025.11		30.6
Gemma-3-4B 2025.11		29.51
LFM2-8B-A1B 2025.11		29.29
Llama-Primus-Merged 2026.01		29.2
ICaRus 2026.02		28.8
ICaRus 2026.02		28.8
Multi Model 2026.02		27.3
IF Model 2026.02		27.2
Multi Model 2026.02		27.2
LFM2-2.6B 2025.11		26.57
Granite-4.0-H 2025.11		26.46
SmolLM3-3B 2025.11		26.31
Llama-3.1-8B-Instruct 2026.01		25.7
FuseLLM 2026.01		24.24
Base Model 2026.02		24.2
Safety 2026.01		23.48
Coding Model 2026.02		21.7
Math Model 2026.02		20.7
Base Model 2026.02		16.7
Base Model 2026.02		16.7