Section Identification on i2b2 Corpus 2010
93.03AccuracyGPT-4
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT-4Prompting strategy=One Shot2024.04 | 93.03 | 100 | 85.36 | 92.1 | 89.45 | |
| Maximum Entropy2024.04 | 91.1 | — | 90.8 | 91 | — | |
| GPT-4Prompting strategy=Zero Shot2024.04 | 89.47 | 100 | 78.46 | 87.93 | 84.58 | |
| LLaMa-2Prompting strategy=Close Ended2024.04 | 88.79 | 100 | 83.57 | 91.05 | 86.54 | |
| GPT-4Prompting strategy=Chain-of-Thought (COT)2024.04 | 86.26 | 99.85 | 74.65 | 85.43 | 84.33 | |
| Regex2024.04 | 66.26 | 100 | 48.27 | 65.11 | 56.8 | |
| LLaMa-2Prompting strategy=Zero Shot2024.04 | 56.2 | 100 | 36.62 | 53.61 | 58.59 | |
| LLaMa-2Prompting strategy=Chain-of-Thought (CoT)2024.04 | 40.23 | 99.83 | 22.61 | 36.87 | 50.7 | |
| MedSpacy2024.04 | 38.45 | 100 | 21.92 | 35.96 | 38.14 | |
| MistralPrompting strategy=Zero Shot2024.04 | 35.3 | 100 | 18.98 | 31.9 | 36.17 | |
| MistralPrompting strategy=Chain-of-Thought (CoT)2024.04 | 32.13 | 99.8 | 17.03 | 29.09 | 33.66 | |
| LLaMa-2Prompting strategy=One Shot2024.04 | 30.54 | 100 | 16.75 | 28.69 | 21.2 | |
| GPT-4Prompting strategy=Close Ended2024.04 | 11.82 | 78.24 | 8.46 | 15.27 | 73.8 | |
| Keyword Based2024.04 | 10.98 | 100 | 8.78 | 16.14 | 69.5 | |
| MistralPrompting strategy=Close Ended2024.04 | 10.41 | 100 | 6.65 | 12.48 | 19.34 | |
| MistralPrompting strategy=One Shot2024.04 | 6.58 | 100 | 3.24 | 6.29 | 7.8 |