Updated with leaderboard benchmark results
Browse files
README.md
CHANGED
|
@@ -8,6 +8,18 @@ This merged model showed marginal improvement in perplexity scores:
|
|
| 8 |
The perplexity for Orca-2-13b is: 7.595028877258301
|
| 9 |
The perplexity for orca2-26B-self-merge is: 7.550178050994873
|
| 10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
|
| 12 |
|
| 13 |
---
|
|
|
|
| 8 |
The perplexity for Orca-2-13b is: 7.595028877258301
|
| 9 |
The perplexity for orca2-26B-self-merge is: 7.550178050994873
|
| 10 |
|
| 11 |
+
### Benchmark Results
|
| 12 |
+
|
| 13 |
+
The following table summarizes the model performance across a range of benchmarks:
|
| 14 |
+
|
| 15 |
+
| Model | Average | ARC | HellaSwag | MMLU | TruthfulQA | Winogrande | GSM8K |
|
| 16 |
+
|------------------------------------|-------------|-------|-----------|-------|------------|------------|-------|
|
| 17 |
+
| microsoft/Orca-2-13b | 58.64 | 60.67 | 79.81 | 60.37 | 56.41 | 76.64 | 17.97 |
|
| 18 |
+
| vmajor/Orca2-13B-selfmerge-26B | 62.24 | 60.84 | 79.84 | 60.32 | 56.38 | 76.87 | 39.2 |
|
| 19 |
+
| vmajor/Orca2-13B-selfmerge-39B | 62.24 | 60.84 | 79.84 | 60.32 | 56.38 | 76.87 | 39.2 |
|
| 20 |
+
|
| 21 |
+
Interestingly the GSM8K performance more than doubled with the first self merge. Second self merge resulting in the 39B model did not produce any further gains.
|
| 22 |
+
|
| 23 |
|
| 24 |
|
| 25 |
---
|