ResearchBenchmarksQuestion Answering on TQ (F1)Follow80.6F1 ScoreIGPO66.97670.51374.0577.587Oct 16, 2025Evaluation ResultsMethodMethodLinksF1 ScoreIGPOBackbone=Qwen2.5-7B-In...Backbone=Qwen2.5-7B-Instruct2025.1080.6GSPOBackbone=Qwen2.5-7B-In...Backbone=Qwen2.5-7B-Instruct2025.1077.7GRPOBackbone=Qwen2.5-7B-In...Backbone=Qwen2.5-7B-Instruct2025.1077PPOBackbone=Qwen2.5-7B-In...Backbone=Qwen2.5-7B-Instruct2025.1075.4RLOOBackbone=Qwen2.5-7B-In...Backbone=Qwen2.5-7B-Instruct2025.1072.5Reinforce++Backbone=Qwen2.5-7B-In...Backbone=Qwen2.5-7B-Instruct2025.1067.5