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

Multitask Language Understanding on MMLU (Accuracy only)

86.3Accuracy

HCP-MAD

Updated 3d ago

Evaluation Results

Method	Links
HCP-MAD 2026.04		86.3
Heter-MAD 2026.04		84.3
DOWN 2026.04		84.06
MAD 2026.04		83.85
Qwen3-32B 2026.04		83.61
SC 2026.04		82.56
SR 2026.04		82.43
SC 2026.04		79.1
MARS 2026.04		78.63
MAD 2026.04		77.95
CoT 2026.04		77.61
SR 2026.04		77.2
Qwen3-8B 2026.04		76.89
CoT 2026.04		76.8
OPD-Qwen3-1.7B 2026.04		74.31
Qwen2.5-7B 2026.04		74.3
Qwen2.5-7B + MathAgent 2026.04		74.1
Qwen2.5-7B + NuminaMath 2026.04		73.8
Qwen2.5-7B + LIMO 2026.04		73.8
Qwen2.5-7B + Magpie 2026.04		73.6
SFT-Qwen3-1.7B 2026.04		73.28
Qwen2.5-7B + S1K 2026.04		73.2
Qwen3-4B 2026.04		72.99
Qwen3-1.7B 2026.04		62.63
DeepSeek V2-Lite 2026.04		57.12
LCA 2026.04		57.04
KDA 2026.04		56.31
FlexPrefill 2026.04		53.85
Qwen3-0.6B 2026.04		52.81
MInference 2026.04		51.01
GPT2-S 2026.04		26.1
GPPoM2-S Hyb. 2026.04		25.6
GPPoM2-S 2026.04		25.5
GPT2-S 2026.04		24.6