Built-in Tasks

EvoToolkit provides several pre-built optimization tasks that demonstrate the power of LLM-driven evolution across different domains.

Available Tasks

Scientific Regression

→ Start Tutorial

Learn how to discover mathematical equations from real scientific datasets.

You'll Learn: - Loading and working with scientific datasets - Creating scientific regression tasks - Using the high-level evotoolkit.solve() API - Comparing different evolutionary algorithms (EoH, EvoEngineer, FunSearch) - Interpreting discovered equations

Prerequisites: Basic Python and NumPy knowledge

---

Prompt Engineering

→ Start Tutorial

Optimize LLM prompts to improve task performance.

You'll Learn: - LLM prompt optimization basics - Using string optimization tasks - Evolving prompt templates - Evaluating and comparing different prompts

Prerequisites: Scientific Regression tutorial

---

Adversarial Attack

→ Start Tutorial

Learn how to evolve adversarial examples and attack algorithms.

You'll Learn: - Creating adversarial attack tasks - Evolving attack strategies - Generating adversarial examples - Evaluating attack effectiveness

Prerequisites: Scientific Regression tutorial, machine learning basics

---

CUDA Tasks

→ Start Tutorial

Optimize GPU kernels using LLM-driven evolution.

You'll Learn: - Creating CUDA optimization tasks - Benchmarking GPU performance - Evolving efficient CUDA kernels - Handling compilation and execution

Prerequisites: CUDA programming basics, GPU hardware

---

Task Comparison

Task	Domain	Difficulty	Best For
Scientific Regression	Data Science	Beginner	Learning the basics, equation discovery
Prompt Engineering	NLP/LLM	Intermediate	Optimizing LLM interactions
Adversarial Attack	Security/ML	Intermediate	Security research, robustness testing
CUDA Tasks	GPU Computing	Advanced	Performance optimization

---

Getting Started

1. Start with Scientific Regression if you're new to EvoToolkit
2. Try Prompt Engineering to see how evolution can optimize text
3. Explore Adversarial Attack for security applications
4. Master CUDA Tasks for GPU optimization

Each tutorial includes complete, runnable code examples that you can adapt for your own problems.