The Red Queen G"odel Machine: Co-Evolving Agents and Their Evaluators

arXiv:2606.26294 · cs.LG, cs.AI, cs.MA, cs.NE · Submitted 2026-06-24 · Read on arXiv

cs.LG, cs.AI, cs.MA, cs.NE

Submitted: 2026-06-24

Updated: 2026-10-08

Comments: 9 pages main text + 31 pages appendix (40 pages total, incl. references); Preliminary preprint; work in progress. Keywords: self-improving agents, learned evaluation, multi-agent systems, auto-mated scientific discovery, controlled utility evolution, co-evolutionary search, autoresearch

Code: https://github.com/SWE-agent/mini-swe-agent

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Benchmark — 3× in this paper · explained in 2 episodes
• Accuracy — 1× in this paper · explained in 1 episode
• Agentic Coding — 1× in this paper · explained in 1 episode
• Papers — 1× in this paper · explained in 3 episodes
• Red Queen Godel Machine — 1× in this paper
• Verifier — 1× in this paper · explained in 2 episodes

Sources

• Hyperagents
• The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search
• DeepSWE: Measuring Frontier Coding Agents on Original, Long-Horizon Engineering Tasks
• Self-Evolving Deep Research via Joint Generation and Evaluation
• Who Grades the Grader? Co-Evolving Evaluation Metrics and Skills for Self-Improving LLM Agents
• APRES: An Agentic Paper Revision and Evaluation System
• Nemotron 3 Super: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning
• Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning
• Live-SWE-agent: Can Software Engineering Agents Self-Evolve on the Fly?
• A Self-Improving Coding Agent
• Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops
• Eliminating Meta Optimization Through Self-Referential Meta Learning
• Arbitrary Order Meta-Learning with Simple Population-Based Evolution
• SELF: Self-Evolution with Language Feedback
• Reflect, Retry, Reward: Self-Improving LLMs via Reinforcement Learning
• Self-Adapting Language Models
• SPICE: Self-Play In Corpus Environments Improves Reasoning
• Language Self-Play For Data-Free Training
• Multi-Agent Evolve: LLM Self-Improve through Co-evolution
• EvoRubrics: Dynamic Rubrics as Rewards via Adversarial Co-Evolution for LLM Reinforcement Learning

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks