ResearchBenchmarksCommonsense Reasoning on HellaSwag 0-shot (test)Follow80.79Accuracy (0-shot)OPPO74.778876.339477.979.4606Sep 30, 2025Evaluation ResultsMethodMethodLinksAccuracy (0-shot)OPPOModel Backbone=Qwen2.5...Model Backbone=Qwen2.5-7B, Evaluation Protocol=0-shot2025.0980.79TRLModel Backbone=Qwen2.5...Model Backbone=Qwen2.5-7B, Evaluation Protocol=0-shot2025.0980.7OPPOModel Backbone=Qwen2.5...Model Backbone=Qwen2.5-3B, Evaluation Protocol=0-shot2025.0975.19TRLModel Backbone=Qwen2.5...Model Backbone=Qwen2.5-3B, Evaluation Protocol=0-shot2025.0975.01