List your company
ResearchFormerly WizWand

Factuality Question Answering on SampleQA (Accuracy %)

3.21Accuracy (%)VRPO
1.41081.87792.3452.8121Aug 5, 2025
Updated 4mo ago

Evaluation results

MethodLinks
VRPOModel Scale=Qwen3-8B2025.083.21
VRPOModel Backbone=Qwen3-8B(Strong Model), Optimization Method=VRPO2025.083.21
Reinforce++Model Scale=Qwen3-8B2025.083.19
Reinforce++Model Backbone=Qwen3-8B(Strong Model), Optimization Method=Reinforce++2025.083.19
KTAEModel Scale=Qwen3-8B2025.083.17
GRPOModel Scale=Qwen3-8B2025.083.03
GRPOModel Backbone=Qwen3-8B(Strong Model), Optimization Method=GRPO2025.083.03
BaseModel Scale=Qwen3-8B2025.082.89
BaseModel Backbone=Qwen3-8B(Strong Model), Optimization Method=Base2025.082.89
PPOModel Scale=Qwen3-8B2025.082.82
VRPOModel Backbone=Qwen2.5-7B-Cold Start(Weak Model), Optimization Method=VRPO2025.082.82
PPOModel Backbone=Qwen3-8B(Strong Model), Optimization Method=PPO2025.082.82
λ-GRPOModel Scale=Qwen3-8B2025.082.77
Reinforce++Model Backbone=Qwen2.5-7B-Cold Start(Weak Model), Optimization Method=Reinforce++2025.082.64
VRPOOptimization Type=test-time optimization, Reward Type=rule-based rewards2025.082.61
PPOModel Backbone=Qwen2.5-7B-Cold Start(Weak Model), Optimization Method=PPO2025.082.54
Dr.GRPOModel Backbone=Qwen2.5-7B-Cold Start(Weak Model), Optimization Method=Dr.GRPO2025.082.54
Dr.GRPOModel Scale=Qwen3-8B2025.082.5
Dr.GRPOModel Backbone=Qwen3-8B(Strong Model), Optimization Method=Dr.GRPO2025.082.5
GRPOOptimization Type=test-time optimization, Reward Type=rule-based rewards2025.082.47
PPOOptimization Type=test-time optimization, Reward Type=rule-based rewards2025.082.43
BaseModel Backbone=Qwen2.5-7B-Cold Start(Weak Model), Optimization Method=Base2025.082.36
GRPOModel Backbone=Qwen2.5-7B-Cold Start(Weak Model), Optimization Method=GRPO2025.082.36
PPOModel Scale=Qwen3-1.7B2025.081.91
BaseModel Scale=Qwen3-1.7B2025.081.78
KTAEModel Scale=Qwen3-1.7B2025.081.64
λ-GRPOModel Scale=Qwen3-1.7B2025.081.62
Cold StartOptimization Type=test-time optimization, Reward Type=rule-based rewards2025.081.62
GRPOModel Scale=Qwen3-1.7B2025.081.6
Dr.GRPOModel Scale=Qwen3-1.7B2025.081.5
VRPOModel Scale=Qwen3-1.7B2025.081.5
Reinforce++Model Scale=Qwen3-1.7B2025.081.48