Update README.md

Author: PeaceRebels

README.md

@@ -60,9 +60,9 @@ That's it! Your agent is now learning and improving. 🎉
 
 AI agents make the same mistakes repeatedly. Fine-tuning is expensive ($1K+ per iteration), slow (days/weeks), and requires labeled data.
 
-**ACE changes that.** Your agents learn from execution feedback—no training data, no fine-tuning, just automatic improvement.
+**ACE changes that.** Based on research from [Stanford & SambaNova](https://arxiv.org/abs/2510.04618), ACE enables agents to learn from execution feedback—no training data, no fine-tuning, just automatic improvement.
 
-ACE agents build a **"playbook"** of strategies that evolve based on experience — learning what works, what doesn't, and continuously improving.
+ACE agents build a **"playbook"** of strategies that evolve based on experience—learning what works, what doesn't, and continuously improving.
 
 ### Clear Benefits
 - 📈 **20-35% Better Performance**: Proven improvements on complex tasks
@@ -135,7 +135,11 @@ The magic happens in the **Playbook**—a living document of strategies that evo
 
 ### The Learning Loop
 
-Task → Execute → Reflect → Curate → Playbook → Better Next Time ↑
+```
+Task → Execute → Reflect → Curate → Playbook → Better Next Time
+  ↑                                                      │
+  └──────────────────────────────────────────────────────┘
+```
 
 Each execution teaches your agent:
 
@@ -144,7 +148,7 @@ Each execution teaches your agent:
 - **🔧 Tool usage** → Discover which tools work best for which tasks
 - **🎯 Edge cases** → Remember rare scenarios and how to handle them
 
-**Key innovation:** All learning happens **in context** through incremental updates—no fine-tuning, no training data, and complete transparency into what your agent learned.
+**Key innovation:** All learning happens **in context** through incremental updates—no fine-tuning, no training data, and complete transparency into what your agent learned. This approach prevents "context collapse" by preserving valuable strategies rather than rewriting the entire playbook.
 
 ```mermaid
 ---
@@ -154,7 +158,7 @@ config:
 ---
 flowchart LR
     Playbook[("`**📚 Playbook**<br>(Evolving Context)<br><br>•Strategy Bullets<br> ✓ Helpful strategies <br>✗ Harmful patterns <br>○ Neutral observations`")]
-    Start(["**📝Query** <br>User prompt or question"]) --> Generator["**⚙️Generator** <br>Creates initial trajectory and answer with tags"]
+    Start(["**📝Query** <br>User prompt or question"]) --> Generator["**⚙️Generator** <br>Executes task using playbook"]
     Generator --> Reflector
     Playbook -. Provides Context .-> Generator
     Environment["**🌍 Task Environment**<br>Evaluates answer<br>Provides feedback"] -- Feedback+ <br>Optional Ground Truth --> Reflector
@@ -165,8 +169,6 @@ flowchart LR
     Generator <--> Environment
 ```
 
-ACE prevents "context collapse" through incremental updates rather than full rewrites, preserving valuable strategies over time.
-
 ---
 
 ## Installation Options