![[2602.23242] A Model-Free Universal AI](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.23242] A Model-Free Universal AI
Summary
This paper presents a groundbreaking model-free agent, AIQI, which achieves asymptotic optimality in reinforcement learning, expanding the landscape of universal agents beyond model-based approaches.
Why It Matters
The introduction of AIQI marks a significant advancement in reinforcement learning by demonstrating that model-free agents can achieve optimal performance. This has implications for AI development, potentially simplifying the design of intelligent systems and broadening their applicability across various domains.
Key Takeaways
- AIQI is the first proven model-free agent to achieve asymptotic ε-optimality in general reinforcement learning.
- The approach utilizes universal induction over distributional action-value functions, differing from traditional policy-based methods.
- Under specific conditions, AIQI is shown to be both asymptotically ε-optimal and ε-Bayes-optimal.
- This research expands the diversity of known universal agents, potentially influencing future AI designs.
- The findings could simplify the development of AI systems by reducing reliance on complex environment models.
Computer Science > Artificial Intelligence arXiv:2602.23242 (cs) [Submitted on 26 Feb 2026] Title:A Model-Free Universal AI Authors:Yegon Kim, Juho Lee View a PDF of the paper titled A Model-Free Universal AI, by Yegon Kim and 1 other authors View PDF Abstract:In general reinforcement learning, all established optimal agents, including AIXI, are model-based, explicitly maintaining and using environment models. This paper introduces Universal AI with Q-Induction (AIQI), the first model-free agent proven to be asymptotically $\varepsilon$-optimal in general RL. AIQI performs universal induction over distributional action-value functions, instead of policies or environments like previous works. Under a grain of truth condition, we prove that AIQI is strong asymptotically $\varepsilon$-optimal and asymptotically $\varepsilon$-Bayes-optimal. Our results significantly expand the diversity of known universal agents. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2602.23242 [cs.AI] (or arXiv:2602.23242v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2602.23242 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Yegon Kim [view email] [v1] Thu, 26 Feb 2026 17:21:16 UTC (141 KB) Full-text links: Access Paper: View a PDF of the paper titled A Model-Free Universal AI, by Yegon Kim and 1 other authorsView PDFTeX Source view license Current browse context: cs.AI < prev | next > new | recent | 2026-02 Change to br...
