Flight Recommendation Evaluation
0.52Inference Time per Round (s)Direct prompting
Evaluation Results
| Method | Links | |
|---|---|---|
| Direct promptingBackbone=Gemma, Calls=1, Hardware=single A100 GPU2026.04 | 0.52 | |
| Direct promptingBackbone=Qwen, Calls=1, Hardware=single A100 GPU2026.04 | 0.63 | |
| Direct promptingBackbone=Llama, Calls=1, Hardware=single A100 GPU2026.04 | 0.74 | |
| Bayesian TeachingBackbone=Qwen, Calls=1, Hardware=single A100 GPU2026.04 | 1.05 | |
| Bayesian TeachingBackbone=Llama, Calls=1, Hardware=single A100 GPU2026.04 | 1.34 | |
| Bayesian TeachingBackbone=Gemma, Calls=1, Hardware=single A100 GPU2026.04 | 1.42 | |
| CoTBackbone=Gemma, Calls=1, Hardware=single A100 GPU2026.04 | 2.15 | |
| CoTBackbone=Qwen, Calls=1, Hardware=single A100 GPU2026.04 | 2.42 | |
| CoTBackbone=Llama, Calls=1, Hardware=single A100 GPU2026.04 | 2.87 | |
| Self-consistencyBackbone=Gemma, Calls=5, Hardware=single A100 GPU2026.04 | 3.67 | |
| Self-consistencyBackbone=Qwen, Calls=5, Hardware=single A100 GPU2026.04 | 3.68 | |
| Self-consistencyBackbone=Llama, Calls=5, Hardware=single A100 GPU2026.04 | 3.79 | |
| ADAPTFUSEBackbone=Gemma, Calls=5, Hardware=single A100 GPU2026.04 | 4.35 | |
| ADAPTFUSEBackbone=Qwen, Calls=5, Hardware=single A100 GPU2026.04 | 4.51 | |
| ADAPTFUSEBackbone=Llama, Calls=5, Hardware=single A100 GPU2026.04 | 4.63 |