Update README.md

README.md

@@ -131,6 +131,8 @@ The model was evaluated on MMLU, ARC-Challenge, GSM-8K, Hellaswag, Winogrande an
 Evaluation was conducted using the Neural Magic fork of [lm-evaluation-harness](https://github.com/neuralmagic/lm-evaluation-harness/tree/llama_3.1_instruct) (branch llama_3.1_instruct) and the [vLLM](https://docs.vllm.ai/en/stable/) engine.
 This version of the lm-evaluation-harness includes versions of MMLU, ARC-Challenge and GSM-8K that match the prompting style of [Meta-Llama-3.1-Instruct-evals](https://huggingface.co/datasets/meta-llama/Meta-Llama-3.1-8B-Instruct-evals).
 
+**Note:** Results have been updated after Meta modified the chat template.
+
 ### Accuracy
 
 #### Open LLM Leaderboard evaluation scores
@@ -148,9 +150,9 @@ This version of the lm-evaluation-harness includes versions of MMLU, ARC-Challen
   <tr>
    <td>MMLU (5-shot)
    </td>
-   <td>69.43
+   <td>68.32
    </td>
-   <td>69.37
+   <td>68.26
    </td>
    <td>99.9%
    </td>
@@ -158,21 +160,21 @@ This version of the lm-evaluation-harness includes versions of MMLU, ARC-Challen
   <tr>
    <td>MMLU (CoT, 0-shot)
    </td>
-   <td>72.56
+   <td>72.83
    </td>
-   <td>72.14
+   <td>72.44
    </td>
-   <td>99.4%
+   <td>99.5%
    </td>
   </tr>
   <tr>
    <td>ARC Challenge (0-shot)
    </td>
-   <td>81.57
+   <td>81.40
    </td>
-   <td>81.48
+   <td>81.40
    </td>
-   <td>99.9%
+   <td>100.0%
    </td>
   </tr>
   <tr>
@@ -180,49 +182,49 @@ This version of the lm-evaluation-harness includes versions of MMLU, ARC-Challen
    </td>
    <td>82.79
    </td>
-   <td>82.64
+   <td>84.31
    </td>
-   <td>99.8%
+   <td>101.8%
    </td>
   </tr>
   <tr>
    <td>Hellaswag (10-shot)
    </td>
-   <td>80.01
+   <td>80.47
    </td>
-   <td>80.1
+   <td>80.07
    </td>
-   <td>100.3%
+   <td>99.5%
    </td>
   </tr>
   <tr>
    <td>Winogrande (5-shot)
    </td>
-   <td>77.90
+   <td>78.06
    </td>
-   <td>77.27
+   <td>77.74
    </td>
-   <td>99.2%
+   <td>99.6%
    </td>
   </tr>
   <tr>
-   <td>TruthfulQA (0-shot, mc2)
+  <td>TruthfulQA (0-shot, mc2)
    </td>
-   <td>54.04
+   <td>54.48
    </td>
-   <td>54.15
+   <td>54.04
    </td>
-   <td>100.2%
+   <td>99.2%
    </td>
   </tr>
   <tr>
    <td><strong>Average</strong>
    </td>
-   <td><strong>74.04</strong>
+   <td><strong>74.05</strong>
    </td>
-   <td><strong>73.89</strong>
+   <td><strong>74.04</strong>
    </td>
-   <td><strong>99.8%</strong>
+   <td><strong>100.0%</strong>
    </td>
   </tr>
 </table>
@@ -235,7 +237,7 @@ The results were obtained using the following commands:
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=3850,max_gen_toks=10,tensor_parallel_size=1 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a16",dtype=auto,max_model_len=3850,max_gen_toks=10,tensor_parallel_size=1 \
   --tasks mmlu_llama_3.1_instruct \
   --fewshot_as_multiturn \
   --apply_chat_template \
@@ -247,7 +249,7 @@ lm_eval \
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=4064,max_gen_toks=1024,tensor_parallel_size=1 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a16",dtype=auto,max_model_len=4064,max_gen_toks=1024,tensor_parallel_size=1 \
   --tasks mmlu_cot_0shot_llama_3.1_instruct \
   --apply_chat_template \
   --num_fewshot 0 \
@@ -258,7 +260,7 @@ lm_eval \
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=3940,max_gen_toks=100,tensor_parallel_size=1 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a16",dtype=auto,max_model_len=3940,max_gen_toks=100,tensor_parallel_size=1 \
   --tasks arc_challenge_llama_3.1_instruct \
   --apply_chat_template \
   --num_fewshot 0 \
@@ -269,7 +271,7 @@ lm_eval \
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=4096,max_gen_toks=1024,tensor_parallel_size=1 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a16",dtype=auto,max_model_len=4096,max_gen_toks=1024,tensor_parallel_size=1 \
   --tasks gsm8k_cot_llama_3.1_instruct \
   --fewshot_as_multiturn \
   --apply_chat_template \