CV

Aleksandar Taranovic

Summary

PhD student in robot learning from human feedback at ALR@KIT, supervised by Prof. Gerhard Neumann.

Education

  • PhD in Computer Science (Robot Learning)
    Karlsruhe Institute of Technology (KIT)
  • M.S. in Electrical Engineering and Computing
    University of Belgrade
  • B.S. in Electrical Engineering and Computing
    University of Belgrade

Work Experience

  • PhD Student, Autonomous Learning Robots (ALR)
    2023-04-01 -
    Karlsruhe Institute of Technology (KIT)
    Research on robot learning from human feedback, supervised by Prof. Gerhard Neumann.
  • PhD Student
    2020-01-01 - 2023-03-31
    Bosch Center for Artificial Intelligence (BCAI)
    Jointly supervised by Dr. Andras Kupcsik and Prof. Gerhard Neumann.
  • Researcher
    -
    Institute of Robotics and Industrial Informatics (IRI), Barcelona
    Research on socially assistive robotics.

Publications

  • Scaffolding Dexterous Manipulation with Vision-Language Models
    2025
    NeurIPS 2025
    Using vision-language models for dexterous robotic manipulation without human demonstrations or handcrafted rewards.
  • AMBER: Adaptive Mesh Generation by Iterative Mesh Resolution Prediction
    2025
    NeurIPS 2025
    A supervised learning approach to adaptive mesh generation that iteratively predicts sizing fields.
  • Acquiring Diverse Skills using Curriculum Reinforcement Learning with Mixture of Experts
    2024
    ICML 2024
    Using curriculum reinforcement learning with mixture of experts to acquire diverse robot skills.
  • Domain-Specific Fine-Tuning of Large Language Models for Interactive Robot Programming
    2024
    ERF 2024
    Fine-tuning large language models for interactive robot programming tasks.
  • Adversarial Imitation Learning with Preferences
    2023
    ICLR 2023
    Extending adversarial imitation learning to simultaneously utilize both demonstrations and preferences.
  • Inferring Versatile Behavior from Demonstrations by Matching Geometric Descriptors
    2022
    CoRL 2022
    Learning versatile behaviors from demonstrations by matching geometric descriptors.
  • Automatic Learning of Cognitive Exercises for Socially Assistive Robotics
    2021
    RO-MAN 2021
    Automatic generation and adaptation of cognitive exercises for socially assistive robots.
  • Adaptable Multimodal Interaction Framework for Robot-Assisted Cognitive Training
    2018
    HRI 2018
    A framework for adjusting robot assistance and interaction modalities in cognitive training.
  • Resource-Based Modality Selection in Robot-Assisted Cognitive Training
    2018
    HRI 2018
    Resource-based approach for selecting interaction modalities in robot-assisted cognitive training.

Interests

  • Robot Learning from Human Feedback
  • Imitation Learning
  • Human-Robot Interaction