Self-Play Only Evolves When Self-Synthetic Pipeline Ensures Learnable Information Gain

Researchers propose a framework for sustainable AI self-evolution through triadic roles (Proposer, Solver, Verifier) that ensures learnable information gain across iterations. The study identifies three key system designs to prevent the common plateau effect in self-play AI systems: asymmetric co-evolution, capacity growth, and proactive information seeking.

• →Current self-play AI systems often plateau quickly because they generate more data without increasing learnable information.
• →Sustainable self-evolution requires three distinct roles: Proposer (task generation), Solver (solution attempts), and Verifier (training signals).
• →Asymmetric co-evolution creates a weak-to-strong-to-weak feedback loop across the three roles.
• →Capacity growth expands computational budgets to match increasing information complexity.
• →Proactive information seeking prevents system saturation by introducing external context and new task sources.