mlsanityData to 7 Oct 2026

A Model-Free Universal AI

arXiv 2602.23242 ↗First posted 26 Feb 2026cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

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. We also apply our novel proof techniques to show asymptotic varepsilon-optimality of Self-AIXI without any ad-hoc assumptions. Our results significantly expand the diversity of known universal agents.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 20 of its 67 references that are in the map. First posted on arXiv 26 Feb 2026.

Builds on

20 of 67 references.

References · 20

  1. 2017Attention Is All You Needcited by 59,932
  2. 2016Asynchronous Methods for Deep Reinforcement Learningcited by 2,218
  3. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,598
  4. 2018World Modelscited by 1,444
  5. 2019Mastering Atari, Go, Chess and Shogi by Planning with a Learned Modelcited by 1,327
  6. 2017A Distributional Perspective on Reinforcement Learningcited by 743
  7. 2015Deep Recurrent Q-Learning for Partially Observable MDPscited by 457
  8. 2020Towards Continual Reinforcement Learning: A Review and Perspectivescited by 268
  9. 2023A Definition of Continual Reinforcement Learningcited by 98
  10. 2016Self-Modification of Policy and Utility Function in Rational Agentscited by 11
  11. 2016Nonparametric General Reinforcement Learningcited by 10
  12. 2019A Strongly Asymptotically Optimal Agent in General Environmentscited by 9
8 more
  1. 2016Thompson Sampling is Asymptotically Optimal in General Environmentscited by 8
  2. 2014Compress and Controlcited by 7
  3. 2025Embedded Universal Predictive Intelligence: a coherent framework for multi-agent learningcited by 7
  4. 2020Curiosity Killed or Incapacitated the Cat and the Asymptotically Optimal Agentcited by 6
  5. 2016A Formal Solution to the Grain of Truth Problemcited by 5
  6. 2019Categorizing Wireheading in Partially Embedded Agentscited by 5
  7. 2014Extreme State Aggregation Beyond MDPscited by 3
  8. 2025Limit-Computable Grains of Truth for Arbitrary Computable Extensive-Form (Un)Known Gamescited by 1

Built on by

Not cited yet.