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

Multitask Language Understanding on MMLU (Average Accuracy)

79.64MMLU Accuracy

FP16

Updated 18d ago

Evaluation Results

Method	Links
FP16 2026.06		79.64
SharQ 2026.06		78.79
NVFP4 2026.06		77.72
FP16 2026.06		74.16
SharQ 2026.06		72.83
NVFP4 2026.06		72.06
FP16 2026.06		65.24
SharQ 2026.06		63.76
MoD-Single Round 2026.06		63.73
Self-Correction 2026.06		63.7
MoD-Multi Round 2026.06		63.69
MoE-LoRA 2026.06		63.59
Multi-Agent Debate 2026.06		63.21
Qwen2.5VL-3b-Instruct 2026.06		62.84
NVFP4 2026.06		61.93
MoD-Multi Round 2026.06		56.35
Self-Correction 2026.06		56.06
Multi-Agent Debate 2026.06		55.97
MoD-Single Round 2026.06		55.95
MoE-LoRA 2026.06		55.62
LLaVA-v1.6-13b 2026.06		55.42
Qwen-VL-Chat 2026.06		50.7
Original Model 2026.06		45.1
Learning to Allocate 2026.06		44.5
TTT+CT-KV 2025.07		44.1
CT-KV 2025.07		43.7
TTT 2025.07		43.6
CT-Prompt 2025.07		43.6
FlexiDepth 2026.06		42.5
AdaSkip 2026.06		41.2
ICL 2025.07		41.2
DoRA 2025.07		40.3
LoRA 2025.07		40.1
Prefix Tuning 2025.07		39.9
Oracle Static Pruning 2026.06		39.8
Prompt Tuning 2025.07		39.2
Rank-Stabilized LoRA 2025.07		38.8
Prefix Tuning 2025.07		38.8
Prompt Tuning 2025.07		37.3
Zero-Shot 2025.07		35.8
Static Pruning 2026.06		34.5