Current interests

My overall goal is to develop systems capable of scientific knowledge discovery. To this end, my current interests revolve around:

  • Machine reasoning: induction, deduction, causation, decomposition, and generalization.
  • Reinforcement learning: active RL settings, goal-directed exploration strategies, implicit RL (i.e., meta-learning approaches to navigating an RL environment), etc.
  • Knowledge representations that enable reasoning, particularly using formal languages, hierarchical, and compositional representations.

Research output

Type
Title
Authors
Year
Publisher
Recognitions
Links
workshopRecursive Decomposition with Dependencies for Generic Divide and Conquer ReasoningSergio Hernández-Gutiérrez, Minttu Alakuijala, Alexander V. Nikitin, Pekka Marttinen2024NeurIPS Sys2 Reasoning
workshopFollowing Ancestral Footsteps: Co-Designing Morphology and Behaviour with Self-Imitation LearningSergio Hernández-Gutiérrez, Ville Kyrki, Kevin S. Luck2024EARL RSS (oral presentation) and EWRLBest Workshop Paper Award (EARL RSS)
thesisSolving Reasoning Problems with Large Language Models via Recursive DecompositionSergio Hernández-Gutiérrez, Pekka Marttinen, Alexander Nikitin, Minttu Alakuijala2024Aalto University
seminarA Comprehensive Overview of Goal-Conditioned Hierarchical Reinforcement Learning: Algorithms, Challenges, and Future DirectionsSergio Hernández-Gutiérrez, Vivienne Wang2023Aalto University
thesisModal Logic Theorem Provers and Validity RatesSergio Hernández-Gutiérrez, Robin Hirsch2019University College London