Vaibhav Singh
PhD Student in Computer Science
Mila, Concordia University · Montreal, Canada
I am a 4th-year PhD student at Mila and Concordia University, advised by my super amazing and motivating advisor, Dr. Eugene Belilovsky. My research focuses on continual learning and efficient inference for large language models.
Previously, I was a Research Intern at ServiceNow, where I worked on efficient diffusion model inference via Mamba-2 backbones. I completed my MS at NYU (working with Dr. Anna Choromanska on EBMs for continual learning), and worked as a Software Engineer at Flipkart.
Research Interests
News
- Oct 2026New preprint on arXiv — RL Post-Training Diffusion Language Models Along Their Own Generation Paths (PACT)
- Sep 2026New preprint on arXiv — Fixed State, Long Reach: What a Constant-Size Cache Buys Block Diffusion at Scale
- Mar 2026Paper accepted at ICML 2026 (main conference, ServiceNow Research Internship) — DiffuMamba: High-Throughput Diffusion LMs with Mamba Backbone
- Jan 2026Paper published in TMLR — Dual-Phase Continual Learning: Supervised Adaptation Meets Unsupervised Retention
- 2025Paper accepted at CoLLAs 2025 (oral) — Beyond Cosine Decay
- 2025Paper at ICML 2025 ESFOMO workshop — Model Parallelism With Subnetwork Data Parallelism
- 2024Paper at CVPR 2024 CLVISION workshop — Wake-Sleep Energy-Based Models for Continual Learning
- Fall 2024TA for Representation Learning (IFT-6150) at Mila
- 2024Research Assistant at Applied AI Institute, Quebec — RAG system for Cybersecurity Ministry
- 2023Started PhD at Mila, Concordia University
Let's Have a Chat
Interested in AI development, research collaboration, or need guidance on building ML systems? Book a free 30-minute consultation with me.
Book a Call