Update README.md
Browse files
README.md
CHANGED
|
@@ -29,6 +29,18 @@ This model is a fine-tuned version of Qwen/Qwen3-8B using the Unsloth framework
|
|
| 29 |
- **Base model:** Qwen/Qwen3-8B
|
| 30 |
- **Fine-tuning method:** LoRA with rank 128
|
| 31 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
## Intended Use
|
| 33 |
|
| 34 |
General conversation, project feedback and brainstorming.
|
|
|
|
| 29 |
- **Base model:** Qwen/Qwen3-8B
|
| 30 |
- **Fine-tuning method:** LoRA with rank 128
|
| 31 |
|
| 32 |
+
## Benchmark Eval
|
| 33 |
+
| Benchmark | Task Category | Accuracy | Raw Score |
|
| 34 |
+
| :--- | :--- | :--- | :--- |
|
| 35 |
+
| **MMLU** | High School Computer Science | 70.0% | 14/20 |
|
| 36 |
+
| **MMLU** | High School Mathematics | 55.0%** | 11/20 |
|
| 37 |
+
| **GSM8K** | Grade School Math Reasoning | **70.0%** | 14/20 |
|
| 38 |
+
| **ARC** | ARC-Challenge (Science) | **15.0%** | 3/20 |
|
| 39 |
+
|
| 40 |
+
### Reproducibility
|
| 41 |
+
Benchmarked using **LabEval**, a lightweight benchmarking suite for M-series Macs and LM Studio.
|
| 42 |
+
[**LabEval on GitHub**](https://github.com/theprint/LabEval)
|
| 43 |
+
|
| 44 |
## Intended Use
|
| 45 |
|
| 46 |
General conversation, project feedback and brainstorming.
|