List your company
ResearchFormerly WizWand

Reading Comprehension on SQuAD v2 (Accuracy)

32.6AccuracyLLaMA3B
19.622.97526.3529.725Oct 14, 2025Nov 20, 2025Dec 28, 2025Feb 4, 2026Mar 13, 2026Apr 20, 2026May 28, 2026
Updated 5mo ago

Evaluation results

MethodLinks
LLaMA3BModel=LLaMA-3.2, Router=None2025.1032.6
Dr.LLMModel=LLaMA-3.2, Router=Enabled2025.1030.6
LLaMA8BModel=LLaMA-3.1, Router=None2025.1029.1
Dr.LLMModel=LLaMA-3.1, Router=Enabled2025.1028.6
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open, Calibration model size=1.7B2026.0527.98
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open2026.0527.56
Qwen3-1.7BType=Base Model2026.0527.19
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open, Calibration model size=4B2026.0527.07
LoRA-NullFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy, Calibration model size=4B2026.0525.84
LoRAFine-tuning dataset=MetaMathQA2026.0525.79
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open, Calibration model size=0.6B2026.0525.53
OPLoRAFine-tuning dataset=MetaMathQA2026.0525.52
CorDAFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy2026.0523.93
CorDAFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy, Calibration model size=4B2026.0523.76
LoRA-NullFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy2026.0523.53
MiLoRAFine-tuning dataset=MetaMathQA2026.0523.07
Qwen3BModel=Qwen2.5, Router=None2025.1021.5
Qwen7BModel=Qwen2.5, Router=None2025.1020.8
Dr.LLMModel=Qwen2.5, Router=Enabled2025.1020.2
Dr.LLMModel=Qwen2.5, Router=Enabled2025.1020.1