Create README.md

README.md

@@ -0,0 +1,13 @@
+FairEval is a human-aligned evaluation framework designed to measure fairness, toxicity, and alignment of generative models using:
+
+LLM-as-Judge scoring
+
+Toxicity + fairness analysis (Detoxify, per-category charts)
+
+Human agreement metrics (κ, ρ)
+
+Group-based fairness dashboard
+
+Model-level SQuAD EM/F1 & uncertainty
+
+This framework is built for rigorous Responsible AI analysis inspired by real-world industry evaluation pipelines.