TinyverseGP: Towards a Modular Cross-domain Benchmarking Framework for Genetic Programming

TL;DR

The paper tackles fragmentation in genetic programming benchmarking across representations and problem domains by introducing TinyverseGP, a modular, cross-domain framework with a lightweight shared library. It defines a unified architecture that supports multiple GP representations and domains, and provides interfaces to established benchmarks such as SRBench, GBFS, and policy-learning environments like Gymnasium. The authors demonstrate integrating a tree-based and graph-based GP across symbolic regression, logic synthesis, and policy search, and emphasize a community-driven, extensible design to facilitate fair, scalable cross-domain comparisons. This framework aims to balance benchmarking rigor with practical usability, enabling rapid adoption and future expansion to additional representations and domains, thereby advancing cross-domain knowledge in GP research.

Abstract

Over the years, genetic programming (GP) has evolved, with many proposed variations, especially in how they represent a solution. Being essentially a program synthesis algorithm, it is capable of tackling multiple problem domains. Current benchmarking initiatives are fragmented, as the different representations are not compared with each other and their performance is not measured across the different domains. In this work, we propose a unified framework, dubbed TinyverseGP (inspired by tinyGP), which provides support to multiple representations and problem domains, including symbolic regression, logic synthesis and policy search.

Figures (3)

• Figure 1: Modular top-level view of TinyverseGP
• Figure 2: High-level architecture of TinyverseGP
• Figure 3: Policy search with TinyverseGP