|
I am an AI resident at Facebook AI Research, advised by Amy Zhang and Yuandong Tian. I'm interested in training universal deep learning models for sequential decision making. Some of my past work has included transformers for reinforcement learning and cross-modal transfer. I previously graduated with a B.S. in Electrical Engineering and Computer Science from UC Berkeley in 2021, where I did undergraduate research as part of the Robot Learning Lab advised by Igor Mordatch and Pieter Abbeel. kzl @ {fb.com, berkeley.edu} / CV / Google Scholar / Twitter / Github |
|
|
|
 
|
Lili Chen*, Kevin Lu*, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas*, Igor Mordatch* Neural Information Processing Systems (NeurIPS), 2021 ICML RL4RealLife Workshop, 2021 (Spotlight Talk) arXiv / website / code SyncedReview article / The Gradient article / Yannic Kilcher video / Eindhoven RL seminar We show that a simple transformer trained with an autoregressive sequence modeling objective can perform competitively with strong specialized offline RL methods. |
 
|
Kevin Lu, Aditya Grover, Pieter Abbeel, Igor Mordatch arXiv preprint, 2021 arXiv / BAIR blog / code The Batch article / VentureBeat article / TWIML podcast / Yannic Kilcher video We show that a transformer pretrained on natural language can, without finetuning of the self-attention and feedforward layers, match the performance of a transformed fully trained on a downstream non-language modality. |
|
|
  |
Michael Laskin*, Denis Yarats*, Hao Liu, Kimin Lee, Albert Zhan, Kevin Lu, Catherine Cang, Lerrel Pinto, Pieter Abbeel Neural Information Processing Systems (NeurIPS), Datasets and Benchmarks Track, 2021 arXiv / code Import AI article We benchmark leading unsupervised exploration algorithms on downstream finetuning performance. |
 
|
Kevin Lu, Aditya Grover, Pieter Abbeel, Igor Mordatch International Conference on Learning Representations (ICLR), 2021 NeurIPS Deep RL Workshop, 2020 (Contributed Talk) ICLR Never-Ending Reinforcement Learning Workshop, 2021 (Invited Paper) arXiv / website / oral / poster / code We show planning over a space of skills is a key component of successful reset-free lifelong learning, avoiding sink states, improving stability, and increasing learning signal. |
 
|
Ruihan Zhao, Kevin Lu, Pieter Abbeel, Stas Tiomkin International Conference on Learning Representations (ICLR), 2021 arXiv / website / code We design a new unbiased empowerment estimator and show it represents empowerment more faithfully than traditional variational mutual information algorithms. |
 
|
Kevin Lu, Igor Mordatch, Pieter Abbeel NeurIPS Deep RL Workshop, 2019 (Contributed Talk) arXiv / website / oral / poster / code Tackling reset-free learning in dynamically changing worlds by combining model-based planning with model-free learning, utilizing an expensive planner only when necessary. |
|
|
|
EECS 126: Probability and Random Processes
CS 70: Discrete Math and Probability |
|
|