Recursive Self-Improvement in AI: Survey of 1,250 Papers with Verification-Strength Taxonomy
A survey of 1,250 arXiv papers (2024–2026) on how AI systems participate in their own improvement. Proposes a verification-strength hierarchy: formal verifiers > process reward models > LLM judges > rubrics > intrinsic self-assessment. Shows improvement quality correlates with this hierarchy. Identifies failure modes including self-confirming loops and model collapse. Flags research direction-setting as the remaining human bottleneck. arXiv:2607.07663.
Why it matters
Comprehensive structured map of a rapidly safety-critical research area. The verification-hierarchy finding has practical design implications: systems using lower-ranked self-evaluation are prone to systematic failure modes. Governance-grade measurement flagged as an underdeveloped area.
Importance: 2/5
1,250-paper comprehensive survey on recursive self-improvement — relevant to model training and safety evaluation design.