Evolutionary AI Research

A comprehensive survey of LLM-powered evolutionary code optimization, autonomous research discovery, and agent framework systems (2024–2026).

50 systems surveyed · By Remigiusz Kinas · MIT License

Evolutionary Systems

LLM-powered evolutionary code optimization frameworks

A-Evolve

Universal Infrastructure for Self-Improving Agents via Agentic Evolution

AB-MCTS / TreeQuest

Adaptive Branching Monte Carlo Tree Search for Multi-LLM Inference-Time Scaling

AI Scientist — Nature Publication

Towards End-to-End Automation of AI Research

Sakana AI: Evolutionary Code Generation for AtCoder Heuristic Contest 058

LLM-Driven Evolutionary Approach to Competitive Optimization Programming

ALE-Bench

A Benchmark for Automated Optimization with LLM-Based Evolutionary Approaches

arXiv Papers: Evolutionary AI for Games & Proofs

Two papers applying LLM-driven evolution to multiagent learning and mathematical theorem proving

AlphaEvolve

A Gemini-Powered Coding Agent for Designing Advanced Algorithms

Arcgentica

Runtime-as-Context Evolutionary Program Synthesis for ARC-AGI-2

AutoEvolver

Can Coding Agents Optimize Algorithms Autonomously?

Confluence Labs: ARC-AGI-2 Solver

State-of-the-Art ARC-AGI-2 Solver via LLM Program Synthesis

Discovering General Methods (DGM)

LLM-Driven Evolutionary Search for General-Purpose Algorithmic Solutions

Imbue Darwinian Evolver for ARC-AGI-2

Evolving Programs Through LLM-Guided Darwinian Natural Selection

DiscoGen

Procedural generator of algorithm discovery tasks spanning 400M+ unique ML problems across 14 domains, enabling meta-meta-learning for evolutionary optimization...

EGGROLL — Evolution Strategies at the Hyperscale

Low-Rank Evolution Strategies Achieving 100x Speedup for Billion-Parameter Model Training

EvoSkill

Self-evolving framework that automatically discovers and refines reusable coding agent skills through iterative failure analysis, Pareto frontier selection, and...

EvoX: Meta-Evolution for Automated Discovery

An adaptive evolution method that optimizes its own evolutionary search process through two-level co-evolution of solutions and search strategies

Evolution Strategies at Scale

First successful application of ES to full-parameter LLM fine-tuning at billion-parameter scale without dimensionality reduction

Evolutionary AI Systems

A Comprehensive Survey of LLM-Powered Evolutionary Code Optimization Frameworks (2024–2026)

GEPA: Optimize Anything

Declarative LLM-Driven Evolutionary Optimization for Text Artifacts

GEPA: Automatically Learning Skills for Coding Agents

Evolutionary Skill Optimization for Repository-Specific Coding Agent Enhancement via GEPA

LLM4AD

Unified Open-Source Platform for LLM-based Automatic Algorithm Design

Matlantis CSP

Crystal Structure Prediction via Genetic Algorithms and Universal Neural Network Potentials

Next Evolution

Architectural Recommendations for the Optimal LLM-Powered Evolutionary System

OpenEvolve

Open-Source Reimplementation of AlphaEvolve for LLM-Guided Evolutionary Coding

Sakana AI: Evolutionary Code Generation for the ICFP Programming Contest 2025

Autonomous LLM-Driven Evolution Applied to Functional Programming Competition Challenges

ShinkaEvolve

Open-Ended Program Evolution Framework

SkyDiscover & AdaEvolve

A Modular Framework for AI-Driven Algorithmic Discovery with Hierarchical Adaptive Search

The AI Scientist

Towards Fully Automated Open-Ended Scientific Discovery

Autoresearch

Autonomous scientific research discovery systems

The AI Scientist v2

End-to-end agentic system that produced the first entirely AI-generated peer-review-accepted workshop paper through progressive tree search over the scientific ...

AI-Researcher

A fully autonomous multi-agent research system that orchestrates the complete scientific pipeline—from literature review and hypothesis generation through algor...

AIRA₂

Asynchronous multi-GPU research agent with evolutionary search, Hidden Consistent Evaluation, and ReAct operators that achieves state-of-the-art on MLE-bench-30...

AutoResearchClaw

Fully autonomous 23-stage pipeline that transforms a research idea into a conference-ready paper with real literature, sandboxed experiments, multi-agent peer r...

Bilevel Autoresearch

A bilevel framework where an outer loop meta-optimizes the inner autoresearch loop by generating and injecting new search mechanisms as Python code at runtime

CycleResearcher

Iterative preference-trained open-source LLM agent pair for full-cycle automated research and peer review via reinforcement learning from reviewer feedback

DeepScientist

Bayesian Optimization-guided autonomous scientific discovery system that surpassed human state-of-the-art on three frontier AI tasks through month-long continuo...

EurekaClaw

Multi-agent AI research assistant that autonomously crawls literature, generates and stress-tests mathematical hypotheses, proves theorems via a 7-stage bottom-...

FARS

First-principles fully automated research system that rejects academic publishing conventions in favor of minimal, composable knowledge units — deployed live wi...

K-Dense BYOK Co-Scientist

Open-source BYOK (Bring Your Own Keys) desktop application that orchestrates a main conversational agent ("Kady") with delegated expert sub-agents equipped with...

Karpathy Autoresearch

Autonomous LLM-driven neural network training research on a single GPU

OpenResearcher

A fully open pipeline for synthesizing long-horizon deep research trajectories via offline corpus bootstrapping and browser-primitive-based browsing

Pi-Autoresearch

Autonomous experiment loop extension for the pi AI coding agent

SkyPilot Scaling Autoresearch

Scaling Karpathy's Autoresearch to Parallel GPU Clusters with Emergent Research Strategies

Zochi

The first AI system to achieve acceptance at an A* scientific conference (ACL 2025), autonomously conducting end-to-end research from literature analysis to pee...

Harness & Agents

Agent frameworks, orchestration, and constraint systems