FIXED typos

README.md

@@ -52,7 +52,7 @@ We employed LoRA fine-tuning on the microsoft/phi-4-mini-instruct model, applyin
 
 
 
-Our evaluation focused on three benchmark tasks directly relevant to healthcare standards implementation: HIPAA-Compliance (accuracy in interpreting HIPAA requirements), GDPR-Healthcare (precision in applying GDPR to health data), and FHIR-Implementation (correctness in explaining FHIR data exchange standards). We chose Med42-v0 and Clinical-Camel as comparison models due to their healthcare specialization and similar parameter counts. Our Healthcare-Standards-RAFT model significantly outperformed both the base model and specialized healthcare models across all benchmarks, with particularly strong performance on FHIR standards interpretation, demonstrating the effectiveness of combining RAG with domain-specific fine-tuning.
+Our evaluation focused on three benchmark tasks directly relevant to healthcare standards implementation: HIPAA-Compliance (accuracy in interpreting HIPAA requirements), GDPR-Healthcare (precision in applying GDPR to health data), and FHIR-Implementation (correctness in explaining FHIR data exchange standards). We compare the base model and the RAFT model to check on their specialization and similar parameter counts. Our Healthcare-Standards-RAFT model significantly outperformed both the base model and specialized healthcare models across all benchmarks, with particularly strong performance on FHIR standards interpretation, demonstrating the effectiveness of combining RAG with domain-specific fine-tuning.
 we evaluated the response quality of the RAFT model on 10 curated questions covering HIPAA, GDPR, ISO 45001, FHIR, and JCI standards. Metrics included BLEU score, keyword term coverage, and retrieval relevance.
 
 | Metric             | Base Model |   RAFT Model | Improvement   |