ResearchBenchmarksAgentic/Tool Use on Agentic/tool BFCL and Tool-2Follow0.711Overall ScoreDASD0.69540.699450.70350.70755May 21, 2026Evaluation ResultsMethodMethodLinksOverall ScoreDASDBackbone=Qwen3-8BBackbone=Qwen3-8B2026.050.711OPSDBackbone=Qwen3-8BBackbone=Qwen3-8B2026.050.703GRPOBackbone=Qwen3-8BBackbone=Qwen3-8B2026.050.701Qwen3-8BPost-training=BasePost-training=Base2026.050.696