|
I am a final year Ph.D. student at MIT CSAIL, advised by Prof. Antonio Torralba. My currently focus on multimodal model, video world model, and reinforcement learning, while I also have research background in physics simulation, optimization, and 3D vision.
Before coming to MIT, I have been fortunate to worked with Prof. Leonidas Guibas and Prof. Gordon Wetzstein at Stanford. I also worked with Prof.Emily Whiting and Prof. Kate Saenko. I am on the industry job market. Reach out if I can contribute.Email: yichenl [at] mit [dot] edu
|
Photo credit: Jiayuan Mao |
|
|
|
|
Nov 23, 2025 Evolutionary algorithm for language model post-training is sample inefficient compared to gradient based approaches, (e.g. traditional RL and SFT). We explore ways to more efficient evolutionary RL, focusing on Low Rank parameter updates and inherent Quantization. |
|
Nov 28, 2025 Autoregressive block diffusion based video generation is prohibitively expensive to rollout efficiently for post-training. We explore ways to densifying reward backpropagation over the sequence length by generating diverse shorter block sequences, compute rewards, then randomly extend a subset for reward recomputation. |
|
|
|
|
| Conference Reviewer: CVPR, ICCV, ECCV, ICML, ICLR,NeurIPS, ACM SIGGRAPH, Journal Reviewer: ACM TOG, IEEE-TPAMI, |
|
|
|
 |
CS231N:Convolutional Neural Networks for Visual Recognition (Spring 2021)
Course Assistant (CA)
|
 |
CS468: Geometric Algorithms: Non-Euclidean Methods (Fall 2020)
Course Assistant (CA)
|
 |
6.S898: Deep Learning (Fall 2023)
Course Assistant (CA)
|
|
|
|
|
|