Update README.md

Author: awinml

README.md

@@ -86,15 +86,15 @@ The API documentation can be found [here](src/llm_blender/README.md).
 
 - A custom component, `LLMBlenderPairRanker`, was developed to integrate the LLM-Blender Framework with Haystack Pipelines. Haystack RAG Pipelines with the LLM-Blender component to ensemble LLMs were evaluated. The pipelines were evaluated on the BillSum and MixInstruct datasets using three metrics: BARTScore, BLEURT, and BERTScore.
 
--We successfully replicated the previously reported results for the LLM-Blender. Moreover, significantly improved performance was observed when utilizing newer LLM models, such as Llama-3-8B, Phi-3-mini and Mistral-7B. These findings demonstrate the potential of ensembling state-of-the-art LLMs to enhance the performance of RAG Pipelines on question-answering, summarization and instruction-following tasks.
+- We successfully replicated the previously reported results for the LLM-Blender. Moreover, significantly improved performance was observed when utilizing newer LLM models, such as Llama-3-8B, Phi-3-mini and Mistral-7B. These findings demonstrate the potential of ensembling state-of-the-art LLMs to enhance the performance of RAG Pipelines on question-answering, summarization and instruction-following tasks.
 
--The authors of LLM-Blender obtained BERTScore values in the range of 62.26 to 74.68 on the MixInstruct dataset. They obtained a BERTScore value of 72.97 with the PairRanker. We obtained BERTScore values in the range of 72.62 to 76.86 using the newer LLMs. We obtained a BERTScore value of 75.83 with the PairRanker ensembling the results from Llama-3-8B, Phi-3-mini and Mistral-7B.
+- The authors of LLM-Blender obtained BERTScore values in the range of 62.26 to 74.68 on the MixInstruct dataset. They obtained a BERTScore value of 72.97 with the PairRanker. We obtained BERTScore values in the range of 72.62 to 76.86 using the newer LLMs. We obtained a BERTScore value of 75.83 with the PairRanker ensembling the results from Llama-3-8B, Phi-3-mini and Mistral-7B.
 
--The authors of LLM-Blender obtained BARTScore values in the range of -4.57 to -3.14 on the MixInstruct dataset. They obtained a BARTScore value of -3.14 with the PairRanker. We obtained BARTScore values in the range of -3.17 to -2.87 using the newer LLMs. We obtained a BARTScore value of -2.87 with the PairRanker ensembling the results from Llama-3-8B, Phi-3-mini and Mistral-7B.
+- The authors of LLM-Blender obtained BARTScore values in the range of -4.57 to -3.14 on the MixInstruct dataset. They obtained a BARTScore value of -3.14 with the PairRanker. We obtained BARTScore values in the range of -3.17 to -2.87 using the newer LLMs. We obtained a BARTScore value of -2.87 with the PairRanker ensembling the results from Llama-3-8B, Phi-3-mini and Mistral-7B.
 
--The authors of LLM-Blender obtained BLEURT values in the range of -1.23 to -0.37 on the MixInstruct dataset. They obtained a BLEURT value of -0.37 with the PairRanker. We obtained BLEURT values in the range of -0.41 to -0.23 using the newer LLMs. We obtained a BLEURT value of -0.26 with the PairRanker ensembling the results from Llama-3-8B, Phi-3-mini and Mistral-7B.
+- The authors of LLM-Blender obtained BLEURT values in the range of -1.23 to -0.37 on the MixInstruct dataset. They obtained a BLEURT value of -0.37 with the PairRanker. We obtained BLEURT values in the range of -0.41 to -0.23 using the newer LLMs. We obtained a BLEURT value of -0.26 with the PairRanker ensembling the results from Llama-3-8B, Phi-3-mini and Mistral-7B.
 
--The newer models like Llama-3-8B, Phi-3-mini, and Mistral-7B significantly outperformed all the models used by the LLM Blender authors on all the three metrics: BERTScore, BARTScore and BLEURT on the MixInstruct dataset.
+- The newer models like Llama-3-8B, Phi-3-mini, and Mistral-7B significantly outperformed all the models used by the LLM Blender authors on all the three metrics: BERTScore, BARTScore and BLEURT on the MixInstruct dataset.
 
 - On the BillSum dataset, we obtained BERTScore values from 73.91 to 75.43, BARTScore values from -3.49 to -3.19, and BLEURT values from -0.39 to -0.20 across the different LLMs. The PairRanker model, ensembling the outputs from Llama-3-8B, Phi-3-mini, and Mistral-7B, achieved the highest scores of 75.83 for BERTScore, -3.19 for BARTScore, and -0.20 for BLEURT.