🤖 AI Summary
Existing LLM-driven program evolution methods predominantly employ overwrite-style mutation, retaining only a single candidate per iteration—leading to loss of beneficial variants, structural degradation, and poor robustness in fragile search spaces. This paper proposes a directed acyclic graph (DAG)-based multi-alternative population representation: nodes store persistent code fragments, and executable programs correspond to paths through the DAG, enabling structural sharing and combinatorial search. We pioneer implicit encoding of quality-diversity optimization into the graph’s topology; further, we introduce substitution-level statistical evaluation and dependency-aware deterministic self-repair, enhancing structural robustness without compromising LLM-generated flexibility. Experiments demonstrate that our method achieves significantly more stable evolution, greater expressive capacity, and superior performance improvement trajectories on program synthesis and meta-learning tasks—outperforming state-of-the-art LLM-guided evolutionary approaches.
📝 Abstract
Large language models (LLMs) are increasingly used to evolve programs and multi-agent systems, yet most existing approaches rely on overwrite-based mutations that maintain only a single candidate at a time. Such methods discard useful variants, suffer from destructive edits, and explore a brittle search space prone to structural failure. We introduce EvoLattice, a framework that represents an entire population of candidate programs or agent behaviors within a single directed acyclic graph. Each node stores multiple persistent alternatives, and every valid path through the graph defines a distinct executable candidate, yielding a large combinatorial search space without duplicating structure. EvoLattice enables fine-grained alternative-level evaluation by scoring each alternative across all paths in which it appears, producing statistics that reveal how local design choices affect global performance. These statistics provide a dense, data-driven feedback signal for LLM-guided mutation, recombination, and pruning, while preserving successful components. Structural correctness is guaranteed by a deterministic self-repair mechanism that enforces acyclicity and dependency consistency independently of the LLM. EvoLattice naturally extends to agent evolution by interpreting alternatives as prompt fragments or sub-agent behaviors. Across program synthesis (proxy and optimizer meta-learning), EvoLattice yields more stable evolution, greater expressivity, and stronger improvement trajectories than prior LLM-guided methods. The resulting dynamics resemble quality-diversity optimization, emerging implicitly from EvoLattice's internal multi-alternative representation rather than an explicit external archive.