Latest News
Deep dives into evolutionary computation, project updates, and community news.
Genesis Tools and Skills: Bringing Project Methods into Development
Custom tools in Genesis 0.13.1 and worktree-aware skill changes in 0.13.4 make project-specific operations and reusable guidance part of the development workflow.
AutoPSO: Evolving the Algorithms That Search
AutoPSO turns particle swarm optimization (PSO) design into a searchable space, showing how EvoX's parallel evaluation supports the evolution of optimization methods themselves.
EvoX Genesis: Building Software That Can Keep Evolving
Genesis extends EvoX's evolutionary AI direction to software: agents generate changes, validation selects accepted work, and the project carries progress forward.
Codex × EvoX: Algorithm Development and Reproducible Experiments
AI coding agents connect research intent to EvoX's executable search workflows, linking problem definition, implementation and reproducible evaluation.
EvoX VirtualES: More Candidates, Fewer Stored Weights
A shared parameter center and per-candidate seeds let VirtualES reconstruct neural-network perturbations on demand, changing the memory budget for population search.
EvoGM: Autonomous LLM Merging via Population Evolution Without Retraining
EvoGM brings evolutionary feedback to generative AI, learning how to combine existing expert-model capabilities without retraining the participating large models.
EvoGP: Scaling the Search for Programs on GPUs
EvoGP makes structured program populations amenable to GPU computation, supporting evolutionary search for interpretable expressions and control policies.
EvoGO: Learning How to Search from Evaluation Feedback
EvoGO connects generative learning with evolutionary optimization, turning evaluated candidates into experience for proposing the next population.
iStratDE: Population Diversity as a Principle for Parallel Search
iStratDE studies how fixed individual-level strategies create diverse search behavior that can be evaluated efficiently with EvoX's GPU computation.
Getting Started with EvoX: Your First Evolutionary Workflow
Build a complete EvoX 1.4.1 experiment and learn how problems, populations, evaluation and execution form a reusable evolutionary workflow.
EvoX Distributed Evaluation: Share the Work, Keep the Search Together
EvoX's experimental distributed workflow partitions candidate evaluation and gathers fitness, extending population search beyond one execution device.
EvoMO: Making Multiobjective Trade-offs Computable at Scale
EvoMO brings tensorized multiobjective search into EvoX, helping experiments explore competing objectives across large candidate populations.