Update README.md

README.md

@@ -25,7 +25,7 @@ language:
 - **Model Developers:** Neural Magic
 
 Quantized version of [Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct).
-It achieves an average score of 72.33 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 73.11.
+It achieves an average score of 73.67 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 74.17.
 
 ### Model Optimizations
 
@@ -162,7 +162,7 @@ oneshot(
 ## Evaluation
 
 The model was evaluated on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) leaderboard tasks (version 1) with the [lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness) and the [vLLM](https://docs.vllm.ai/en/stable/) engine, using the following command.
-A modified version of ARC-C was used for evaluations, in line with Llama 3.1's prompting.
+A modified version of ARC-C and GSM8k-cot was used for evaluations, in line with Llama 3.1's prompting. It can be accessed on the [Neural Magic fork of the lm-evaluation-harness](https://github.com/neuralmagic/lm-evaluation-harness/tree/llama_3.1_instruct).
 ```
 lm_eval \
   --model vllm \
@@ -206,13 +206,13 @@ lm_eval \
    </td>
   </tr>
   <tr>
-   <td>GSM-8K (5-shot, strict-match)
+   <td>GSM-8K-cot (8-shot, strict-match)
    </td>
-   <td>75.66
+   <td>82.03
    </td>
-   <td>73.77
+   <td>81.80
    </td>
-   <td>97.50%
+   <td>99.72%
    </td>
   </tr>
   <tr>
@@ -248,11 +248,11 @@ lm_eval \
   <tr>
    <td><strong>Average</strong>
    </td>
-   <td><strong>73.11</strong>
+   <td><strong>74.17</strong>
    </td>
-   <td><strong>72.33</strong>
+   <td><strong>73.67</strong>
    </td>
-   <td><strong>98.94%</strong>
+   <td><strong>99.33%</strong>
    </td>
   </tr>
 </table>