mlsanityData to 14 Oct 2026

MAGIK: Mapping to Analogous Goals via Imagination-enabled Knowledge Transfer

arXiv 2506.01623 ↗First posted 2 Jun 2025cs.AI cs.LGcited by 1

On the map: LLM reasoning → Reinforcement learning for reasoning

Abstract (arXiv)

Humans excel at analogical reasoning - applying knowledge from one task to a related one with minimal relearning. In contrast, reinforcement learning (RL) agents typically require extensive retraining even when new tasks share structural similarities with previously learned ones. In this work, we propose MAGIK, a novel framework that enables RL agents to transfer knowledge to analogous tasks without interacting with the target environment. Our approach leverages an imagination mechanism to map entities in the target task to their analogues in the source domain, allowing the agent to reuse its original policy. Experiments on custom MiniGrid and MuJoCo tasks show that MAGIK achieves effective zero-shot transfer using only a small number of human-labelled examples. We compare our approach to related baselines and highlight how it offers a novel and effective mechanism for knowledge transfer via imagination-based analogy mapping.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Reinforcement learning for reasoning, one of 17 directions in LLM reasoning. Builds on 28 of its 35 references that are in the map. First posted on arXiv 2 Jun 2025.

Builds on

28 of 35 references.

References · 28

  1. 2018Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actorcited by 4,885
  2. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,739
  3. 2017FiLM: Visual Reasoning with a General Conditioning Layercited by 2,333
  4. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,653
  5. 2018World Modelscited by 1,513
  6. 2018Learning Latent Dynamics for Planning from Pixelscited by 1,264
  7. 2016The Concrete Distribution: A Continuous Relaxation of Discrete Random Variablescited by 1,148
  8. 2014Semi-Supervised Learning with Deep Generative Modelscited by 496
  9. 2018Quantifying Generalization in Reinforcement Learningcited by 352
  10. 2016Successor Features for Transfer in Reinforcement Learningcited by 340
  11. 2023Minigrid & Miniworld: Modular & Customizable Reinforcement Learning Environments for Goal-Oriented Taskscited by 327
  12. 2018Visual Reinforcement Learning with Imagined Goalscited by 251
16 more
  1. 2018A Study on Overfitting in Deep Reinforcement Learningcited by 174
  2. 2017DARLA: Improving Zero-Shot Transfer in Reinforcement Learningcited by 149
  3. 2018Universal Successor Features Approximatorscited by 103
  4. 2019Transfer in Deep Reinforcement Learning Using Successor Features and Generalised Policy Improvementcited by 88
  5. 2022Does Zero-Shot Reinforcement Learning Exist?cited by 83
  6. 2016Deep Successor Reinforcement Learningcited by 72
  7. 2017Learning Disentangled Representations with Semi-Supervised Deep Generative Modelscited by 72
  8. 2016Deep Reinforcement Learning with Successor Features for Navigation across Similar Environmentscited by 50
  9. 2018Transfer Learning for Related Reinforcement Learning Tasks via Image-to-Image Translationcited by 50
  10. 2022Semi-Supervised Offline Reinforcement Learning with Action-Free Trajectoriescited by 28
  11. 2016Generalizing Skills with Semi-Supervised Reinforcement Learningcited by 18
  12. 2024Learning Successor Features the Simple Waycited by 10
  13. 2024Distributional Successor Features Enable Zero-Shot Policy Optimizationcited by 8
  14. 2021Semi-Supervised Disentanglement of Class-Related and Class-Independent Factors in VAEcited by 7
  15. 2023EMOTE: An Explainable architecture for Modelling the Other Through Empathycited by 1
  16. 2024Dynamic Policy Fusion for User Alignment Without Re-Interactioncited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026ASPECT:Analogical Semantic Policy Execution via Language Conditioned Transfercited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Compositional Transduction with Latent Analogies for Offline Goal-Conditioned Reinforcement Learning
  2. 2023Foundations for Transfer in Reinforcement Learning: A Taxonomy of Knowledge Modalities
  3. 2023Enhancing data efficiency in reinforcement learning: a novel imagination mechanism based on mesh information propagation
  4. 2023Dream to Adapt: Meta Reinforcement Learning by Latent Context Imagination and MDP Imagination
  5. 2025Atomic Skills are the Prerequisite: When Reinforcement Learning Synthesizes Compositional Reasoning, and When It Only Amplifies
  6. 2026Prism: Policy Reuse via Interpretable Strategy Mapping in Reinforcement Learning
  7. 2026EVOKE: Eliciting World Knowledge in Agents for Transferable Decision-Making
  8. 2025Beyond 'Aha!': Toward Systematic Meta-Abilities Alignment in Large Reasoning Models