avatar

Niko Lorantos

University of Massachusetts Amherst
nlorantos (at) umass (dot) edu


I study mathematics and philosophy at UMass Amherst. My research focuses on open-endedness and adaptive systems. Previously I built coding agent infrastructure for ecosystem model calibration at Woodwell Climate, designed evolutionary algorithms and artificial life with Lee Spector at Amherst College, engineered language models at UMass CICS, and developed structural biology models at Neoclease.

In my free time I enjoy snowboarding, music, and spending time outdoors.

Available for full-time research engineering starting mid-2027.

Research

  1. RLC
    Toward Aligned Open-Endedness in AI Systems

    Niko Lorantos

    Reinforcement Learning Conference Finding the Frame Workshop 2026
    We show that alignment is incompatible with open-endedness when defined over the output space of a generative system and identify specific demands open-endedness places beyond existing alignment methods.
    PDF

  2. arXiv
    Adaptive Exploration in Lenia with Intrinsic Multi-Objective Ranking

    Niko Lorantos, Lee Spector

    arXiv
    We evolve populations of Lenia artificial life through intrinsic objectives to promote diversity, emergent complexity, and open-endedness.
    PDF arXiv Code

  3. GPTP XXII
    Evolution of Artificial Intelligence, Continued

    Lee Spector, Ryan Bahlous-Boldi, Paul Vander Vort, Niko Lorantos

    Genetic Programming Theory & Practice XXII
    We outline significant advancements in AI driven by evolution, particularly in neural architecture design and preference learning.
    Book

  4. IC2S2
    Estimating Support for US Presidential Candidates Using X Polls

    Chhandak Bagchi, Gayathri Sunil, Arash Khojaste, Vishal Kalakonnavar, Niko Lorantos, Mattia Samory, Junghwan Yang, Przemyslaw Grabowicz

    12th International Conference on Computational Social Science