ResearchBenchmarksCommonsense Question Answering on ECQAFollow57.2Performance Score (Finetune Baseline vs Predict Baseline)T5-base42.22446.1125053.888May 4, 2023Evaluation ResultsMethodMethodLinksPerformance Score (Finetune Baseline vs Predict Baseline)Performance Score (Finetune Baseline vs Predict Infusion)Simulatability ScorePerformance Score (Finetune Infusion vs Predict Infusion)TREU ScoreAccuracyT5-baseBackbone=T5-baseBackbone=T5-base2023.0557.274.617.498.959.1—BART-baseBackbone=BART-baseBackbone=BART-base2023.0542.843.8190.148.3—InfoRidgeBackbone=Qwen-2.5-0.5BBackbone=Qwen-2.5-0.5B2025.07—————65.45InfoRidgeBackbone=LLaMA-3.1-8BBackbone=LLaMA-3.1-8B2025.07—————80.49