List your company
ResearchFormerly WizWand

Factual QA on NQ-Open

48.34AccuracyCorVer
3.651215.253126.85538.4569May 28, 2026
Updated 5mo ago

Evaluation results

MethodLinks
CorVerModel=Llama-3.1-8B2026.0548.34
CorVerModel=Llama-3.2-3B2026.0543.41
FoRAGModel=Llama-3.1-8B2026.0542.71
RLFHModel=Llama-3.1-8B2026.0542.44
RawModel=Llama-3.1-8B2026.0540.66
FSPOModel=Llama-3.1-8B2026.0539.25
KnowRLModel=Llama-3.1-8B2026.0538.53
FoRAGModel=Llama-3.2-3B2026.0537.15
RLFHModel=Llama-3.2-3B2026.0535.9
RawModel=Llama-3.2-3B2026.0534.13
CorVerModel=Qwen3-8B2026.0532.8
KnowRLModel=Llama-3.2-3B2026.0530.53
FoRAGModel=Qwen3-8B2026.0529.81
FSPOModel=Qwen3-8B2026.0529.67
RawModel=Qwen3-8B2026.0529.61
RLFHModel=Qwen3-8B2026.0529.42
KnowRLModel=Qwen3-8B2026.0528.45
CorVerModel=Qwen3-4B2026.0526.59
FSPOModel=Qwen3-4B2026.0525.65
FoRAGModel=Qwen3-4B2026.0525.43
RLFHModel=Qwen3-4B2026.0525.12
RawModel=Qwen3-4B2026.0524.65
KnowRLModel=Qwen3-4B2026.0517.3
FSPOModel=Llama-3.2-3B2026.0510.61
Qwen3-1.7BType=Base Model2026.058.81
OPLoRAFine-tuning dataset=MetaMathQA2026.057.97
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open2026.057.73
CorDAFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy, Calibration model size=4B2026.057.5
LoRA-NullFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy, Calibration model size=4B2026.057.45
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open, Calibration model size=4B2026.057.25
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open, Calibration model size=1.7B2026.057
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open, Calibration model size=0.6B2026.056.34
LoRAFine-tuning dataset=MetaMathQA2026.056.06
CorDAFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy2026.055.73
MiLoRAFine-tuning dataset=MetaMathQA2026.055.4
LoRA-NullFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy2026.055.37