diff --git a/README.md b/README.md
index 697c436..3105dc1 100644
--- a/README.md
+++ b/README.md
@@ -5,7 +5,7 @@
 [![arXiv](https://img.shields.io/badge/arXiv-2603.24511-b31b1b.svg?logo=arxiv)](https://arxiv.org/abs/2603.24511)
 
 <p align="center">
-  <img src="assets/teaser.png" width="90%" alt="Claude autoresearch vs Optuna hyperparameter search: best train and validation loss over trials">
+  <img src="assets/pareto_evolution_small.png" width="90%" alt="Pareto front with evolution annotations">
 </p>
 
 We show that an *[autoresearch](https://github.com/karpathy/autoresearch)*-style pipeline powered by Claude Code discovers novel white-box adversarial attack *algorithms* that **significantly outperform** all existing [methods](claudini/methods/original/README.md) in jailbreaking and prompt injection evaluations.
diff --git a/assets/pareto_evolution_small.png b/assets/pareto_evolution_small.png
new file mode 100644
index 0000000..35127a5
Binary files /dev/null and b/assets/pareto_evolution_small.png differ
diff --git a/assets/teaser.png b/assets/teaser.png
deleted file mode 100644
index 49c479d..0000000
Binary files a/assets/teaser.png and /dev/null differ