Update README.md
Browse files
README.md
CHANGED
@@ -122,7 +122,12 @@ Next, we evaluate **METAGENE-1** on the GUE multi-species classification benchma
|
|
122 |
| **COVID** | 22.2 | 23.3 | 62.2 | **73.0** | 71.9 | 72.5 |
|
123 |
| **Global Win %** | 0.0 | 0.0 | 7.1 | 21.4 | 25.0 | **46.4** |
|
124 |
|
|
|
|
|
|
|
|
|
|
|
125 |
|
126 |
## **Model Details**
|
127 |
-
- **Release Date**:
|
128 |
- **Model License**: Apache 2.0
|
|
|
122 |
| **COVID** | 22.2 | 23.3 | 62.2 | **73.0** | 71.9 | 72.5 |
|
123 |
| **Global Win %** | 0.0 | 0.0 | 7.1 | 21.4 | 25.0 | **46.4** |
|
124 |
|
125 |
+
## **Safety Considerations**
|
126 |
+
|
127 |
+
- **Potential Risks**: METAGENE-1 is a metagenomic model designed for biosurveillance and anomaly detection, but future larger variants could pose risks if misused for synthetic biology tasks, such as designing pathogenic sequences.
|
128 |
+
- **Purpose and Capabilities**: Unlike models aimed at long-sequence generation, METAGENE-1 is optimized for detecting anomalies in short metagenomic reads (100-300 base pairs) and is limited by its 512-token context length, minimizing design misuse.
|
129 |
+
- **Open Source Impact**: We believe open-sourcing METAGENE-1 will positively impact research in pathogen monitoring, but we advocate for stricter safety evaluations before releasing more capable future models.
|
130 |
|
131 |
## **Model Details**
|
132 |
+
- **Release Date**: Jan 06 2025
|
133 |
- **Model License**: Apache 2.0
|