Update README.md

README.md

@@ -35,7 +35,7 @@ configs:
 
 This dataset is a human-annotated subset of randomly sampled question-answer entries from the CommonsenseQA dataset (tau/commonsense_qa).
 The 'rationales' for each QA pair were created using a two-part method. First, Mixtral (mistralai/Mixtral-8x7B-Instruct-v0.1) was used
-to generate 3 unique CoT (Chain-of-Thought) explanations. Next, human evaluation was applied to distill the randomly sampling down to
+to generate 3 unique CoT (Chain-of-Thought) explanations. Next, human evaluation was applied to distill the random sampling down to
 a cohesive set of question-answer-rationale triplets.
 
 The working hypothesis, inspired by the research papers listed below, is that a diverse set of CoT rationales passed along