Antonis Antoniades

PhD Student, Computer Science, UCSB

prof_pic.jpg
Henley Hall
Santa Barbara
California

While studying Physics at the University of California, Santa Barbara, I became interested in the connection between biological and artificial intelligence, and in particular how they could interact across multiple levels of abstraction. My research led me to pursue a PhD in Computer Science at UCSB, co-advised by Dr. William Wang (CS, UCSB NLP Group) and Dr. Spencer LaVere Smith (SLAB Neuroscience & Neuroengineering Lab).

My current research focuses include:

  1. Open-ended systems, with emphasis on developing robust dynamic planning approaches and exploring self-assembling capabilities that can adapt across diverse tasks.

  2. Bio-inspired and cognitive approaches to machine learning, particularly investigating neuromorphic computing architectures and collective intelligence.

  3. Understanding fundamental principles that enable complex skills in Large Language Models, including exploring their generalization boundaries and implicit vs. explicit reasoning capabilities.

In terms of industry, I particularly enjoyed my internship at Leela AI, working on multi-agent reinforcement learning. I also lead the development of Calibrex, a product aiming to democratize data-driven resistance training.

Another big part of who I am is a guitar player 🎸 - which is probably what I am most talented at. Some of my other interests include swimming, surfing, gaming, photography, reading, meditation and chess.

Selected Publications

  1. SWE-Search_demo.gif
    SWE-Search: Enhancing Software Agents with Monte Carlo Tree Search and Iterative Refinement
    Antonis Antoniades, Albert Örwall, Kexun Zhang, and 3 more authors
    2024
  2. multiagent_attack.png
    MultiAgent Collaboration Attack: Investigating Adversarial Attacks in Large Language Model Collaborations via Debate
    Alfonso Amayuelas, Xianjun Yang, Antonis Antoniades, and 3 more authors
    In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2024
  3. generate_neurons.gif
    Neuroformer: Multimodal and Multitask Generative Pretraining for Brain Data
    Antonis Antoniades, Yiyi Yu, Joseph Canzano, and 2 more authors
    In The Twelfth International Conference on Learning Representations, 2024

News

Oct 25, 2024 Presented our research on neural generative models at the Stanford Machine Learning and Neurotheory Journal Club.
Sep 1, 2024 Our work on generating human-guided counterfactual explanations for molecular property prediction has been accepted at KDD 2024. Really enjoyed being in Barcelona during the summer. 😁
May 3, 2024 Delivered a talk on generative pretraining for brain data at UC Santa Barbara’s Machine Learning Journal Club.
Apr 16, 2024 Gave an invited presentation on neural data modeling approaches at the University of Washington Computational Neuroscience Journal Club.
Apr 11, 2024 Presented our recent work on multimodal generative modeling for neural data at the Caltech Neuroscience and Machine Learning Journal Club.