Recent advances in generative AI are reshaping how we design and discover mathematical and scientific artifacts. Problems in mathematics, engineering, and the sciences have recently been solved by systems that use LLMs to evolve search and optimization algorithms. If you can specify what you are looking for, they can help you find it.
Concretely, LLM-guided evolution can help find abstract objects representable in a computer — vectors, matrices, networks, programs, or collections thereof. It must be possible to write code that reports whether an object meets the desired requirements. Often, LLMs can write this code for us.
This workshop gives researchers a concentrated, hands-on introduction to these tools, focused on ShinkaEvolve, an open-source framework available for participants to continue using after the workshop.
Gain practical fluency with LLM-guided evolutionary search through guided tutorials, mentorship, and full infrastructure support from YCRC.
Take a problem from idea to working prototype: formulate optimization problems, improve search systems, or develop new benchmarks.
Work alongside researchers from mathematics, CS, biology, materials science, and beyond. Make connections and find new collaborators.
Share what you built, reflect on what worked, suggest improvements, discuss how LLM-guided evolution can improve our research.