↓Skip to main content

🏢 University of British Columbia

Propensity Score Alignment of Unpaired Multimodal Data
·2058 words·10 mins· loading · loading
AI Generated Multimodal Learning Vision-Language Models 🏢 University of British Columbia
Unlocking multimodal learning’s potential with propensity scores: This novel approach aligns unpaired data across modalities, significantly improving representation learning.
Local Superior Soups: A Catalyst for Model Merging in Cross-Silo Federated Learning
·3305 words·16 mins· loading · loading
AI Generated Machine Learning Federated Learning 🏢 University of British Columbia
Local Superior Soups (LSS) significantly accelerates federated learning by efficiently merging pre-trained models, drastically cutting communication rounds without sacrificing accuracy.
Leveraging Environment Interaction for Automated PDDL Translation and Planning with Large Language Models
·1918 words·10 mins· loading · loading
Natural Language Processing Large Language Models 🏢 University of British Columbia
This paper presents a fully automated method for PDDL translation and planning using LLMs and environment interaction, achieving a 66% success rate on challenging PDDL domains.
Implicit Optimization Bias of Next-token Prediction in Linear Models
·1645 words·8 mins· loading · loading
Natural Language Processing Large Language Models 🏢 University of British Columbia
Researchers reveal implicit optimization biases in next-token prediction for language models, showing how gradient descent selects solutions based on data sparsity and a novel margin concept, impactin…
Heavy-Tailed Class Imbalance and Why Adam Outperforms Gradient Descent on Language Models
·3990 words·19 mins· loading · loading
Large Language Models 🏢 University of British Columbia
Adam’s superior performance on language models stems from its resilience to heavy-tailed class imbalance, unlike SGD, which struggles with infrequent word losses.
General bounds on the quality of Bayesian coresets
·1364 words·7 mins· loading · loading
AI Theory Optimization 🏢 University of British Columbia
New theoretical bounds on Bayesian coreset approximation errors enable efficient large-scale Bayesian inference, overcoming prior limitations and improving coreset construction methods.
Even Sparser Graph Transformers
·2059 words·10 mins· loading · loading
Machine Learning Deep Learning 🏢 University of British Columbia
Spexphormer achieves significant memory reduction in graph Transformers by leveraging a two-stage training process that leverages attention score consistency across network widths to effectively spars…
ET-Flow: Equivariant Flow-Matching for Molecular Conformer Generation
·2269 words·11 mins· loading · loading
AI Applications Healthcare 🏢 University of British Columbia
ET-Flow, a novel equivariant flow-matching model, generates highly accurate and physically realistic molecular conformers significantly faster than existing methods.
Adaptive Randomized Smoothing: Certified Adversarial Robustness for Multi-Step Defences
·3521 words·17 mins· loading · loading
Image Classification 🏢 University of British Columbia
Adaptive Randomized Smoothing certifies deep learning model predictions against adversarial attacks by cleverly combining randomized smoothing with adaptive, multi-step input masking for improved accu…
3D Gaussian Splatting as Markov Chain Monte Carlo
·1616 words·8 mins· loading · loading
3D Vision 🏢 University of British Columbia
Researchers rethink 3D Gaussian Splatting as MCMC sampling, improving rendering quality and Gaussian control via a novel relocation strategy.