Dialogue Strategy Alignment on AnnoMI (full)
3.4MI-iGPT3.5-ICL-ALL
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT3.5-ICL-ALLStrategy Category=W/o Strategies, Base Model=GPT-3.5, ICL Type=ALL2024.03 | 3.4 | 6.25 | 1.53 | 28.4 | 70.6 | |
| GPT3.5Strategy Category=W/o Strategies, Base Model=GPT-3.52024.03 | 3.1 | 44.8 | 1.29 | 43 | 70.6 | |
| GPT4-ICL-KNNStrategy Category=W/o Strategies, Base Model=GPT-4, ICL Type=KNN2024.03 | 3.1 | 4.18 | 1.06 | 45.8 | 71.8 | |
| GPT4-ICL-RANDStrategy Category=W/o Strategies, Base Model=GPT-4, ICL Type=RAND2024.03 | 3 | 10.55 | 1.01 | 45.1 | 71.8 | |
| GPT3.5-ICL-RANDStrategy Category=W/o Strategies, Base Model=GPT-3.5, ICL Type=RAND2024.03 | 2.8 | 7.05 | 0.98 | 51.7 | 82.3 | |
| GPT4-ICL-ALLStrategy Category=W/o Strategies, Base Model=GPT-4, ICL Type=ALL2024.03 | 2.4 | 10.8 | 0.91 | 38 | 64.8 | |
| GPT3.5-ICL-KNNStrategy Category=W/o Strategies, Base Model=GPT-3.5, ICL Type=KNN2024.03 | 2.1 | 4.28 | 0.93 | 54.3 | 82.8 | |
| GPT4Strategy Category=W/o Strategies, Base Model=GPT-42024.03 | 1.4 | 11.8 | 0.86 | 41.1 | 71.3 | |
| GoldDescription=Human Ground Truth2024.03 | 1.4 | 2.3 | 1.51 | 59 | 81.7 | |
| GPT3.5-DIIRStrategy Category=With Strategies (Ours), Base Model=GPT-3.5, Method=DIIR2024.03 | 1.3 | 115 | 0.99 | 61.1 | 95.1 | |
| GPT4-DIIRStrategy Category=With Strategies (Ours), Base Model=GPT-4, Method=DIIR2024.03 | 0.6 | 135 | 1.04 | 59.6 | 94.4 |