Question Answering on HaluEval (Hallucination, Accuracy)
69AccuracyCAROL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CAROLModel=GPT–5–nano, Latency (s)=1.062 ± 0.015, Cost ($)=0.0163 ± 0.0012026.05 | 69 | 13.8 | |
| RAGModel=GPT–5–nano, Latency (s)=0.517 ± 0.007, Cost ($)=0.0376 ± 0.0012026.05 | 63.8 | — | |
| CAROLModel=Llama–3.1–8B, Latency (s)=1.630 ± 0.0202026.05 | 61.6 | 25.4 | |
| RAGModel=Llama–3.1–8B, Latency (s)=1.096 ± 0.0102026.05 | 57.4 | — |