Commit
•
1dd84f8
1
Parent(s):
e2325db
Update README.md
Browse files
README.md
CHANGED
|
@@ -23,17 +23,6 @@ Llama-Spark is intended for use in conversational AI applications, such as chatb
|
|
| 23 |
## Training Information
|
| 24 |
|
| 25 |
Llama-Spark is built upon the Llama-3.1-8B base model, fine-tuned using of the Tome Dataset and merged with Llama-3.1-8B-Instruct.
|
| 26 |
-
## Evaluation Results
|
| 27 |
-
We have removed our initial benchmark results and replaced them with the results from the OpenLLM Leaderboard, which are much more deterministic:
|
| 28 |
-
| Metric |Value|
|
| 29 |
-
|-------------------|----:|
|
| 30 |
-
|Avg. |24.90|
|
| 31 |
-
|IFEval (0-Shot) |79.11|
|
| 32 |
-
|BBH (3-Shot) |29.77|
|
| 33 |
-
|MATH Lvl 5 (4-Shot)| 1.06|
|
| 34 |
-
|GPQA (0-shot) | 6.60|
|
| 35 |
-
|MuSR (0-shot) | 2.62|
|
| 36 |
-
|MMLU-PRO (5-shot) |30.23|
|
| 37 |
|
| 38 |
## Acknowledgements
|
| 39 |
|
|
|
|
| 23 |
## Training Information
|
| 24 |
|
| 25 |
Llama-Spark is built upon the Llama-3.1-8B base model, fine-tuned using of the Tome Dataset and merged with Llama-3.1-8B-Instruct.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
|
| 27 |
## Acknowledgements
|
| 28 |
|