ExLab

Perceptual grounding and explainable AI lab in Informatics at the University of Edinburgh.

Our work broadly involves foundational research at the intersection of machine learning, natural language processing, computer vision, and cognitive science. Topics of research include explainability and interpretability for perception, multimodal generative AI, structured / graph-based representation learning and reasoning, Bayesian program learning, human-like perception and reasoning, effective variational inference, and probabilistic programming.


People



news

May 6, 2025 Our paper on Bernoulli Priors for Diffusion Autoencoders was accepted to the CVPR 2025 Workshop on Generative Models for Computer Vision.
May 1, 2025 Our paper on Banyan: improved embeddings with structure was accepted to ICML 2025.
Oct 10, 2024 Our paper on Are LLMs good pragmatic speakers? was accepted to the NeurIPS 2024 Workshop on Behavioral ML.
May 10, 2024 Three papers accepted to ICML on autoencoding CNPs, library learning by decompiling knowledge, and better learning with sinusoidal position embeddings!
May 2, 2024 Our paper on graph kernel convolutions for interpretable classification was accepted to the DMLR workshop at ICLR.

selected publications

2025

  1. banyan.png
    Banyan: Improved Representation Learning with Explicit Structure
    Mattia Opper, and N. Siddharth
    In International Conference on Machine Learning (ICML), Jul 2025

2024

  1. ddecomp.png
    DreamDecompiler: Improved Bayesian Program Learning by Decompiling Amortised Knowledge
    Alessandro B. Palmarini, Christopher G. Lucas, and N. Siddharth
    In International Conference on Machine Learning (ICML), Jul 2024
  2. ppsvae.png
    Autoencoding Conditional Neural Processes for Representation Learning
    Victor Prokhorov, Ivan Titov, and N. Siddharth
    In International Conference on Machine Learning (ICML), Jul 2024

2023

  1. strae.png
    StrAE: Autoencoding for Pre-Trained Embeddings using Explicit Structure
    Mattia Opper, Victor Prokhorov, and N. Siddharth
    In Empirical Methods in Natural Language Processing (EMNLP), Dec 2023

2022

  1. dood.png
    Drawing out of Distribution with Neuro-Symbolic Generative Models
    Yichao Liang, Joshua B Tenenbaum, Tuan Anh Le, and N Siddharth
    In Advances in Neural Information Processing Systems (NeurIPS), Dec 2022
  2. meme.png
    Learning multimodal VAEs through mutual supervision
    Tom Joy, Yuge Shi, Philip H. S. Torr, Tom Rainforth, and 2 more authors
    In International Conference on Learning Representations (ICLR), May 2022