Adding Evaluation Results
#3
by
leaderboard-pr-bot
- opened
README.md
CHANGED
|
@@ -6,4 +6,17 @@ This model was the result of a 50/50 average weight merge between Airoboros-33B-
|
|
| 6 |
After prolonged testing we concluded that while this merge is highly flexible and capable of many different tasks, it has to much variation in how it answers to be reliable.
|
| 7 |
Because of this the model relies on some luck to get good results, and is therefore not recommended to people seeking a consistent experience, or people sensitive to anticipation based addictions.
|
| 8 |
|
| 9 |
-
If you would like an improved version of this model that is more stable check out my Airochronos-33B merge.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
After prolonged testing we concluded that while this merge is highly flexible and capable of many different tasks, it has to much variation in how it answers to be reliable.
|
| 7 |
Because of this the model relies on some luck to get good results, and is therefore not recommended to people seeking a consistent experience, or people sensitive to anticipation based addictions.
|
| 8 |
|
| 9 |
+
If you would like an improved version of this model that is more stable check out my Airochronos-33B merge.
|
| 10 |
+
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
| 11 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_Henk717__chronoboros-33B)
|
| 12 |
+
|
| 13 |
+
| Metric | Value |
|
| 14 |
+
|-----------------------|---------------------------|
|
| 15 |
+
| Avg. | 51.45 |
|
| 16 |
+
| ARC (25-shot) | 63.91 |
|
| 17 |
+
| HellaSwag (10-shot) | 85.0 |
|
| 18 |
+
| MMLU (5-shot) | 59.44 |
|
| 19 |
+
| TruthfulQA (0-shot) | 49.83 |
|
| 20 |
+
| Winogrande (5-shot) | 80.35 |
|
| 21 |
+
| GSM8K (5-shot) | 15.01 |
|
| 22 |
+
| DROP (3-shot) | 6.62 |
|