readme

README.md

@@ -37,7 +37,6 @@ conda create -n wildbench python=3.10
 conda activate wildbench
 pip install vllm -U # pip install -e vllm 
 pip install openai datasets tenacity
-# pip install google-cloud-aiplatform 
 pip install google-generativeai
 pip install cohere mistralai 
 pip install anthropic==0.19.0
@@ -111,11 +110,11 @@ You should change the code to add these APIs, for example, gemini, cohere, claud
         </ul>
         We use three reference models (GPT-4-turbo-0429, Claude-3-Haiku, and Llama-2-70B-chat) to compute the rewards for each model. The final WB Reward-Mix is the average of the three rewards on 1024 examples.
         <h4>Mitigating Length Bias</h4>  
-        As many studies have shown, LLM judges tend to prefer longer responses. To mitigate this bias, we propose a simple and customizable length penalty method. <b>We convert Slightly Win/Lose to be a Tie if the winner is longer than the loser by a certain length threshold (K characters).</b> We set K=500 by default, but you can customize it on our leaderboard UI. Note that <b>K= ∞ will disable the length penalty.</b>
+        As many studies have shown, LLM judges tend to prefer longer responses. To mitigate this bias, we propose a simple and customizable length penalty method. <b>We convert Slightly Win/Lose to be a Tie if the winner is longer than the loser by a certain length threshold (K characters).</b>  Note that <b>K= ∞ will disable the length penalty.</b>
     </div>
 </details>
 
-### Run evaluation scripts 
+### ‼️ Run evaluation scripts 
 
 We suggest to use OpenAI's [Batch Mode](https://platform.openai.com/docs/guides/batch) for evaluation, which is faster, cheaper and more reliable.