New README format

.gitignore

@@ -1,2 +1,3 @@
 /readme-generator.toml
 /kl-results.csv
+/.DS_Store

README.md

@@ -23,7 +23,9 @@ base_model_relation: quantized
 
 ![KL divergence versus language-model checkpoint size](assets/kl-vs-size.svg)
 
-KL divergence is shown on a logarithmic scale. Size is the decimal GB occupied by the language-model checkpoint; vision encoders, MTP heads, and speculative-decoding models are excluded.
+Mirai-M is on the size-KL Pareto frontier: we found no checkpoint that is smaller while also having lower KL divergence.
+
+*Evaluation data mixture: 45% public agentic, 30% public SFT/long-context, 25% private chat data.*
 
 ### Quickstart