Tag
Introduces Flex, a new DSPy module that lets language models rewrite the program code itself rather than just prompts, enabling better optimization, fewer model calls, and safer execution via sandboxing.
Introduces HERO, an LLM-based program optimizer that overcomes the weakest-link effect by generating and recombining heterogeneous atomic edits, achieving faster convergence and higher scores across algorithmic, game, agentic, and robotic domains.
An exploration of superoptimization techniques for finding the smallest possible program, reviewing related research and implementations.
This paper proposes Retrieval Augmented Search (RAS), a blackbox adaptation method using retrieval-augmented search and atomic edit decomposition (AEGIS) to improve LLM-based program optimization for C++ and Python, achieving up to 2.06x improvement over prior methods.