Lexington Whalen
NVIDIA Research
I am a Research Scientist on the DLER team in Santa Clara, CA. Previously, at SoftBank's SB Intuitions, I co-led pretraining and inference optimization for the Sarashina LLM family — Japan's largest language models. Before that, during my first stint at NVIDIA Research, I co-developed Nemotron-Labs-Diffusion (EMNLP 2026) and co-led Efficient-DLM (ICML 2026). I was an NSF Graduate Research Fellow and pursued an M.S. in CS at Georgia Tech. I also serve as an informal advisor to government officials, and have previously worked with representatives in California and embassy members in Japan.
Note: I can only advise on topics that are public knowledge, and can only give my opinion based on publicly available materials.
I am a history enthusiast and enjoy discussing history, meaning, and geopolitics. I also enjoy climbing and bouldering, and am trying to get better at it. There's other stuff too probably.
- Nov 2026Looking for a nice bouldering club in Santa Clara or SF! Please feel free to let me know if you have any suggestions!
- Oct 2026Rejoining NVIDIA Research as a Research Scientist on the DLER team in Santa Clara, CA! Starting November 16.
- Oct 2026Talk at SB Intuitions: Faster, Smaller, Lighter: Best Practices for LLM Efficiency!
- Sep 2026New tech report: KBBQ: A Predictive Noise Law and the Limits of Spectrum Flattening in FP4 Quantization!
- Sep 2026Joined the U.S. Embassy Tokyo roundtable Inspiring the Next Phase of U.S.–Japan Technology Cooperation: Physical AI! Great to meet leaders from industry, academia, and government across the U.S. and Japan, including Noetra.
- Aug 2026Nemotron-Labs-Diffusion accepted at EMNLP 2026 (Main Conference)!
- Jul 2026Talk at SB Intuitions: Diffusion Language Models: From Fundamentals to Applications!
- Jun 2026SB Intuitions releases Sarashina2.2-TTS, a SoTA Japanese TTS model, and the Sarashina3 mini/nano tech report. For the bigger models, we'll be using our in-house spec decoding recipe!
- Jun 2026New preprint: Teaching Diffusion to Speculate Left-to-Right — optimizations to raise accepted draft length by up to 76%!
- Jun 2026Nemotron-Labs-Diffusion released on Hugging Face — ~1M total downloads, 6× tokens-per-pass over Qwen3-8B!
- Jun 2026Efficient-DLM accepted at ICML 2026!
- Jun 2026NeRArch-Sim accepted at ISCA 2026!
- Apr 2026Joined SoftBank SB Intuitions as Pretraining & Inference Optimization Research Lead. Gave the SoftBank Group entrance ceremony speech!
- Apr 2025LAMB accepted at ACL 2025!
- Apr 2025Early-Bird Diffusion accepted at CVPR 2025!