Jack Lu
Hey! I’m a CS PhD candidate in the CILVR lab at NYU Courant, advised by Mengye Ren and supported by the NSERC PGS-D. Previously, I did my undergrad in CS + Math at UWaterloo.
I study reasoning [1, 2, 3, 4] and adaptation [5, 6] in LLMs and embodied agents, as well as world models [7, 8] that agents can learn and plan in. I’m currently a visiting researcher at Meta working on recursive self-improvement, and this summer I worked on reasoning VLAs and world models at NVIDIA. Before these, I did research and engineering for autonomous driving and ML for health at NVIDIA, Waabi/Uber-ATG, IBM, and DarwinAI, working with Raquel Urtasun, Sanja Fidler, and Alexander Wong.
I’m happy to discuss collaboration, mentorship, and research in general. You can email me for a virtual or in-person chat. My office is at 60 5th Ave, New York.
news
| Sep 2026 | Started as a visiting researcher at Meta, working on recursive self-improvement! |
|---|---|
| Sep 2026 | The Surprising Effectiveness of Deleting Weights in LLM Post-Training is accepted by NeurIPS 2026, and Solaris is accepted as a NeurIPS 2026 spotlight. See ya in 🇦🇺 |
| Jul 2026 | HyperThink: Text-to-Parameter Hypernetworks for Efficient Reasoning is accepted by COLM 2026. See ya in 🌉 |
| May 2026 | Arrived at NVIDIA Santa Clara to work on VLA and world models. 10/10 ☀️, 0/10 walkability. |
| May 2026 | The Surprising Effectiveness of Deleting Weights in LLM Reasoning and Adaptation is accepted by the ICML 2026 Workshop on Foundations of Deep Generative Models. |
| May 2026 | Context Tuning for In-Context Optimization is accepted by ICML 2026. See ya in 🇰🇷 |
| Jan 2026 | When Does Verification Pay Off? A Closer Look at LLMs as Solution Verifiers is accepted by the ICLR 2026 Workshop on AI with Recursive Self-Improvement and featured by the NYU Center for Data Science. |
| Jan 2026 | SkillFactory: Self-Distillation For Learning Cognitive Behaviors is accepted by ICLR 2026. See ya in 🇧🇷 |
| Jul 2024 | ProCreate, Don’t Reproduce! Propulsive Energy Diffusion for Creative Generation is accepted by ECCV 2024. See ya in 🇮🇹 |
| Apr 2024 | Awarded the NSERC PGS-D Scholarship to support my PhD at NYU. |
| Jan 2024 | SceneControl: Diffusion for Controllable Traffic Scene Generation is accepted by ICRA 2024. |
| Sep 2023 | Started my PhD in Computer Science at NYU, advised by Mengye Ren. |
publications
- Preprint
TacitVLA: Reasoning Vision-Language-Action Models Without Reasoning LatencyUnder review, 2026
