Question Answering on TruthfulQA (Hallucination, Accuracy)
55.6AccuracyCAROL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CAROLModel=GPT–5–nano, Latency (s)=164.8 ± 1.35, Cost ($)=0.1463 ± 0.0032026.05 | 55.6 | 25.7 | |
| CAROLModel=Llama–3.1–8B, Latency (s)=194.3 ± 1.602026.05 | 54 | 34.8 | |
| RAGModel=GPT–5–nano, Latency (s)=119.2 ± 1.10, Cost ($)=0.2933 ± 0.0042026.05 | 35.6 | — | |
| RAGModel=Llama–3.1–8B, Latency (s)=120.9 ± 1.252026.05 | 12.7 | — |