Tool-calling on DuplexSLA-Bench tool-call
91.33Avg Accuracy (3 patterns)ASR + LLM sys
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| ASR + LLM sysType=Cascade baseline (Streaming ASR + LLM)2026.05 | 91.33 | 2.77 | 89.33 | 2.33 | 89.33 | 4.71 | 95.33 | 1.27 | |
| DuplexSLA2026.05 | 85.56 | 0.64 | 85.67 | 0.67 | 75 | 0.68 | 96 | 0.57 |