ResearchBenchmarksCross-domain Generalization on Cross-Domain Summary Code Scientific Agentic toolFollow70Average ScoreDASD68.02468.53769.0569.563May 21, 2026Evaluation ResultsMethodMethodLinksAverage ScoreDASDBackbone=Qwen3-8BBackbone=Qwen3-8B2026.0570GRPOBackbone=Qwen3-8BBackbone=Qwen3-8B2026.0568.5Qwen3-8BPost-training=BasePost-training=Base2026.0568.1OPSDBackbone=Qwen3-8BBackbone=Qwen3-8B2026.0568.1