Natural Language Understanding on NLU++
29.07Subset AccuracyAdaptive ToR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Adaptive ToRLatency (s)=9.73, Ave Depth=1.53, LLM Calls=6.012026.04 | 29.07 | 71.79 | 71.51 | |
| Standard RAGLatency (s)=5.61, Ave Depth=0.00, LLM Calls=2.02026.04 | 27.29 | 63.16 | 67.71 | |
| ToR-RAGLatency (s)=15.58, Ave Depth=3.00, LLM Calls=10.542026.04 | 26.51 | 70.97 | 71.43 |