List your company
ResearchFormerly WizWand

Factual question answering on TriviaQA

76.52AccuracyCorVer
9.003226.531644.0661.5884Apr 17, 2026Apr 23, 2026Apr 30, 2026May 7, 2026May 14, 2026May 21, 2026May 28, 2026
Updated 5mo ago

Evaluation results

MethodLinks
CorVerModel=Llama-3.1-8B2026.0576.52
RawModel=Llama-3.1-8B2026.0571.86
RLFHModel=Llama-3.1-8B2026.0569.55
FoRAGModel=Llama-3.1-8B2026.0569.4
KnowRLModel=Llama-3.1-8B2026.0568.41
CorVerModel=Qwen3-8B2026.0563.99
FSPOModel=Llama-3.1-8B2026.0563.43
RawModel=Qwen3-8B2026.0562.84
CorVerModel=Llama-3.2-3B2026.0562.24
FSPOModel=Qwen3-8B2026.0561.81
RLFHModel=Qwen3-8B2026.0561.39
FoRAGModel=Qwen3-8B2026.0561.34
KnowRLModel=Qwen3-8B2026.0560.92
FoRAGModel=Llama-3.2-3B2026.0560.66
RLFHModel=Llama-3.2-3B2026.0560.66
RawModel=Llama-3.2-3B2026.0555.39
CorVerModel=Qwen3-4B2026.0553.77
FSPOModel=Qwen3-4B2026.0552.88
FoRAGModel=Qwen3-4B2026.0552.65
RLFHModel=Qwen3-4B2026.0552.5
RawModel=Qwen3-4B2026.0551.14
KnowRLModel=Llama-3.2-3B2026.0549.6
KnowRLModel=Qwen3-4B2026.0540.9
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open, Calibration model size=1.7B2026.0532.46
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open2026.0532.41
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open, Calibration model size=0.6B2026.0532.18
FoLoRAFine-tuning dataset=MetaMathQA, Calibration data=NQ Open, Calibration model size=4B2026.0532.07
LoRA-NullFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy, Calibration model size=4B2026.0530.58
LoRA-NullFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy2026.0530.53
MiLoRAFine-tuning dataset=MetaMathQA2026.0529.69
LoRAFine-tuning dataset=MetaMathQA2026.0529.62
OPLoRAFine-tuning dataset=MetaMathQA2026.0529.59
Qwen3-1.7BType=Base Model2026.0528.78
NAG_SmolLM3-3BTarget Setting=Single-Target, Backbone=SmolLM3-3B2026.0422.6
NAG_Qwen3-1.7BTarget Setting=Single-Target, Backbone=Qwen3-1.7B2026.0422.3
FSPOModel=Llama-3.2-3B2026.0522.28
NAG_Llama-3.2-3BTarget Setting=Single-Target, Backbone=Llama-3.2-3B2026.0421.3
BETRTarget Setting=Single-Target2026.0420.2
FineWeb-Edu2026.0420.1
NAG_SmolLM3-3BTarget Setting=Multi-Target, Backbone=SmolLM3-3B2026.0419.9
CorDAFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy, Calibration model size=4B2026.0519.85
NAG_Qwen3-1.7BTarget Setting=Multi-Target, Backbone=Qwen3-1.7B2026.0419.2
NAG_Llama-3.2-3BTarget Setting=Multi-Target, Backbone=Llama-3.2-3B2026.0418
CorDAFine-tuning dataset=MetaMathQA, Calibration data=model-generated pretraining-proxy2026.0516.58
Random2026.0415.6
BETRTarget Setting=Multi-Target2026.0411.6