I’m an researcher currently at OpenAI. I graduated from UC Berkeley in 2021 where I worked with Pieter Abbeel and Igor Mordatch on reinforcement learning and sequence modeling.

Email: matches my arxiv papers

Note: my Twitter account is @_kevinlu; all other accounts are fake.

Quick highlights:

Blogs

• Mar 2024 — Spending Inference Time
• Feb 2024 — LoRAs as Composable Programs
• Feb 2024 — Unifying RLHF Objectives

Papers

• Jun 2021 — Decision Transformer
• Mar 2021 — Pretrained Transformers as Universal Computation Engines