Update README.md
Browse files
README.md
CHANGED
|
@@ -46,7 +46,7 @@ DNA inputs are broken into overlapping 6‑mer tokens and padded or truncated to
|
|
| 46 |
outputs use the LLaMA tokenizer, augmented with `<DNA>` as a special token to mark sequence insertion points.
|
| 47 |
|
| 48 |
### Limitations and Disclaimer
|
| 49 |
-
ChatNT can only handle questions related to the 27 tasks it has been trained on. ChatNT is **not** a clinical or diagnostic tool.
|
| 50 |
It can produce incorrect or “hallucinated” answers, particularly on out‑of‑distribution inputs, and its numeric predictions may suffer digit‑level errors. Confidence
|
| 51 |
estimates require post‑hoc calibration. Users should always validate critical outputs against experiments or specialized bioinformatics
|
| 52 |
pipelines.
|
|
|
|
| 46 |
outputs use the LLaMA tokenizer, augmented with `<DNA>` as a special token to mark sequence insertion points.
|
| 47 |
|
| 48 |
### Limitations and Disclaimer
|
| 49 |
+
ChatNT can only handle questions related to the 27 tasks it has been trained on, including the same format of DNA sequences. ChatNT is **not** a clinical or diagnostic tool.
|
| 50 |
It can produce incorrect or “hallucinated” answers, particularly on out‑of‑distribution inputs, and its numeric predictions may suffer digit‑level errors. Confidence
|
| 51 |
estimates require post‑hoc calibration. Users should always validate critical outputs against experiments or specialized bioinformatics
|
| 52 |
pipelines.
|