ResearchBenchmarksTool Use on Tool Use Domain AverageTool Use on Tool Use Domain AverageFollow71.1Tool Use Average AccuracyDASD69.5469.94570.3570.755May 21, 2026Evaluation resultsMethodMethodLinksTool Use Average AccuracyDASDBackbone=Qwen3-8BBackbone=Qwen3-8B2026.0571.1OPSDBackbone=Qwen3-8BBackbone=Qwen3-8B2026.0570.3GRPOBackbone=Qwen3-8BBackbone=Qwen3-8B2026.0570.1BaseBackbone=Qwen3-8BBackbone=Qwen3-8B2026.0569.6