Tool Use on tau2-bench (productivity, proactivity, personalization)
43ProductivityMCPE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MCPEAgent=GPT-5-mini, User=GPT-4o, Communication Mode=CPE-optimized2026.06 | 43 | 7.5 | 63 | |
| MCPEAgent=GPT-5-mini, User=Qwen3-VL-32B, Communication Mode=CPE-optimized2026.06 | 38 | 14.5 | 60.3 | |
| MCPEAgent=DeepSeek-V3.2, User=GPT-4o, Communication Mode=CPE-optimized2026.06 | 30 | 0.5 | 52 | |
| MCPEAgent=DeepSeek-V3.2, User=Qwen3-VL-32B, Communication Mode=CPE-optimized2026.06 | 30 | 5.5 | 57.5 |