About me

I’m a Research Scientist at Samsung AI Lab Montreal, leading research on graph neural networks, large language models and optimization, with applications in language reasoning and scientific discovery.

News

• Aug 2025: (Almost) Free Modality Stitching of Foundation Models accepted to EMNLP 2025 (Main Conference)
• May 2025: Celo: Training Versatile Learned Optimizers on a Compute Diet accepted to TMLR (tmlr)
• Apr 2025: Gave an invited talk at the ICLR 2025 Workshop on Weight Space Learning, see slides and talk
• Mar 2025: Meta-learning Optimizers for Communication-Efficient Learning accepted to TMLR (tmlr)
• Feb 2025: Became Adjunct Professor at the University of Montreal

• Jan 2025: NiNo accepted at ICLR 2025

• Dec 2024: Invited as a speaker at ICLR 2025 Workshop on Weight Space Learning
• Oct 2024: µLO accepted as oral at OPT for ML 2024 NeurIPS Workshop, congrats Benjamin and Charles-Étienne!
• Feb 2024: Neural Graphs accepted as oral at ICLR 2024, congrats Miltos!

Reviewing

• 2024: ICML, NeurIPS, ICLR, Neural Networks
• 2023: ICML, NeurIPs, MLG @ KDD, ICLR
• 2022: CVPR, ICML, ICML Workshop, NeurIPS, Learning on Graphs Conference (LoG), Nature Machine Intelligence
• 2021: ICCV (Outstanding Reviewer, top 5% student reviewers)

Selected publications

See the full list at Google Scholar.

• Accelerating training with neuron interaction and nowcasting networks
  Boris Knyazev, Abhinav Moudgil, Guillaume Lajoie, Eugene Belilovsky, Simon Lacoste-Julien
  International Conference on Learning Representations (ICLR), 2025
  openreview, arxiv, code, poster, twitter

• Graph Neural Networks for Learning Equivariant Representations of Neural Networks
  Miltiadis Kofinas, Boris Knyazev, Yan Zhang, Yunlu Chen, Gertjan J Burghouts, Efstratios Gavves, Cees GM Snoek, David W Zhang
  International Conference on Learning Representations (ICLR), 2024 (oral)
  openreview, arxiv, code, twitter

• Can We Scale Transformers to Predict Parameters of Diverse ImageNet Models?
  Boris Knyazev, Doha Hwang, Simon Lacoste-Julien
  International Conference on Machine Learning (ICML), 2023
  arxiv, video, code

• Hyper-Representations as Generative Models: Sampling Unseen Neural Network Weights
  Konstantin Schürholt, Boris Knyazev, Xavier Giró-i-Nieto, Damian Borth
  Advances in Neural Information Processing Systems (NeurIPS), 2022
  arxiv, slides, code

• On Evaluation Metrics for Graph Generative Models
  Rylee Thompson, Boris Knyazev, Elahe Ghalebi, Jungtaek Kim, Graham W Taylor
  International Conference on Learning Representations (ICLR), 2022
  openreview, arxiv, code

• Parameter Prediction for Unseen Deep Architectures
  Boris Knyazev, Michal Drozdzal, Graham W. Taylor, Adriana Romero-Soriano
  Advances in Neural Information Processing Systems (NeurIPS), 2021
  openreview, arxiv, UofG news, Yannic Kilcher’s video, neurips video, code, Colab-predict, Colab-fine-tune, twitter, quantamagazine

• Brick-by-Brick: Combinatorial Construction with Deep Reinforcement Learning
  Hyunsoo Chung, Jungtaek Kim, Boris Knyazev, Jinhwi Lee, Graham W. Taylor, Jaesik Park, Minsu Cho
  Advances in Neural Information Processing Systems (NeurIPS), 2021
  openreview, arxiv, video, code

• Context-aware Scene Graph Generation with Seq2Seq Transformers
  Yichao Lu, Himanshu Rai, Jason Chang, Boris Knyazev, Shashank Shekhar, Graham W. Taylor, Maksims Volkovs
  International Conference on Computer Vision (ICCV), 2021
  openaccess, pdf, code

• Generative Compositional Augmentations for Scene Graph Prediction
  Boris Knyazev, Harm de Vries, Cătălina Cangea, Graham W. Taylor, Aaron Courville, Eugene Belilovsky
  International Conference on Computer Vision (ICCV), 2021
  arxiv, ICML Workshop version, ICML workshop video, code

• Graph Density-Aware Losses for Novel Compositions in Scene Graph Generation
  Boris Knyazev, Harm de Vries, Cătălina Cangea, Graham W. Taylor, Aaron Courville, Eugene Belilovsky
  British Machine Vision Conference (BMVC), 2020
  arxiv, bmvc, code, Data Fest tutorial

• Learning Temporal Attention in Dynamic Graphs with Bilinear Interactions
  Boris Knyazev*, Carolyn Augusta*, Graham Taylor (*equal contribution)
  PLOS ONE, 2021
  arxiv, plos one journal link, code

• Understanding Attention and Generalization in Graph Neural Networks
  Boris Knyazev, Graham Taylor, Mohamed Amer
  Advances in Neural Information Processing Systems (NeurIPS), 2019
  arxiv, neurips, ICLR Workshop version, code, poster, slides

• Image Classification with Hierarchical Multigraph Networks
  Boris Knyazev, Xiao Lin, Mohamed Amer, Graham Taylor
  British Machine Vision Conference (BMVC), 2019
  arxiv, bmvc pdf, code, blog post

Blog posts

• Optimizing LLMs Faster by Learning Connections: Neuron Interaction and Nowcasting Networks

• Tutorial on Graph Neural Networks for Computer Vision and Beyond

• Anisotropic, Dynamic, Spectral and Multiscale Filters Defined on Graphs

• Spectral Graph Convolution Explained and Implemented Step By Step

• Can we do better than Convolutional Neural Networks?

Open source contributions

Geometric Deep Learning Extension Library for PyTorch

Last updated: Sep, 2025