Commit
·
8e05648
1
Parent(s):
a0a497f
Update README.md
Browse files
README.md
CHANGED
|
@@ -41,7 +41,22 @@ ical data modalities via a single GPT model. BioMedGPT aligns different biologic
|
|
| 41 |
|
| 42 |
**Intended Use Cases**
|
| 43 |
|
| 44 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
|
| 46 |
**Out-of-scope Uses**
|
| 47 |
|
|
|
|
| 41 |
|
| 42 |
**Intended Use Cases**
|
| 43 |
|
| 44 |
+
| **Method** | Parameters (B) | Setting | MedMCQA(\%) | PubMedQA(\%) |
|
| 45 |
+
|------------------------|----------------|-----------|-------------|--------------|
|
| 46 |
+
| Human (pass)* | - | Manual | - | 60.0 |
|
| 47 |
+
| Human (expert)* | - | Manual | 90 | 78.0 |
|
| 48 |
+
|------------------------|----------------|-----------|-------------|--------------|
|
| 49 |
+
| InstructGPT* | 175 | zero-shot | 44.0 | 73.2 |
|
| 50 |
+
| ChatGPT* | - | zero-shot | 44.7 | 63.9 |
|
| 51 |
+
| Llama* | 7 | zero-shot | 24.3 | 5.2 |
|
| 52 |
+
| Llama2 | 7 | zero-shot | 30.6 | 3.7 |
|
| 53 |
+
| Llama2-Chat | 7 | zero-shot | 35.5 | 21.9 |
|
| 54 |
+
|------------------------|----------------| --------- |-------------|--------------|
|
| 55 |
+
| Llama | 7 |Fine-tuing | 48.2 | 73.4 |
|
| 56 |
+
| Llama2-Chat | 7 |Fine-tuing | 48.3 | 75.5 |
|
| 57 |
+
| PMC-Llama | 7 |Fine-tuing | 50.5 | 69.5 |
|
| 58 |
+
|------------------------|----------------|-----------|-------------|--------------|
|
| 59 |
+
| BioMed-Llama2 | 7 |Fine-tuing | **51.4** | **76.1** |
|
| 60 |
|
| 61 |
**Out-of-scope Uses**
|
| 62 |
|