Minqi Jiang

London, United Kingdom

Minqi Jiang is a PhD student at UCL DARK and Meta AI, where he is advised by Professors Tim Rocktäschel and Edward Grefenstette. He is primarily interested in how decision-making models can be trained in simulation to transfer robustly and safely to the real world. His recent works focus on how adaptive curricula can be used to produce more robust deep reinforcement learning agents. By kickstarting an open-ended co-evolutionary process between agent and environment, such curricula are promising avenues to more generally capable AI. He was previously the co-founder of a startup that provided human-in-the-loop automation to the travel industry and the product manager for Google Translate.

Minqi's Publications

May 05, 2022

CORE MACHINE LEARNING

Prioritized Level Replay

Minqi Jiang, Edward Grefenstette, Tim Rocktäschel

May 05, 2022

March 09, 2022

CORE MACHINE LEARNING

Replay-Guided Adversarial Environment Design

Minqi Jiang, Michael Dennis, Jack Parker-Holder, Jakob Foerster, Edward Grefenstette, Tim Rocktäschel

March 09, 2022

February 16, 2022

RESEARCH

Evolving Curricula with Regret-Based Environment Design

Minqi Jiang, Jack Parker-Holder, Michael Dennis, Mikayel Samvelyan, Jakob Foerster, Edward Grefenstette, Tim Rocktäschel

February 16, 2022

August 22, 2013

RESEARCH

WordCraft: An Environment for Benchmarking Commonsense Agents

Minqi Jiang, Jelena Luketina, Nantas Nardelli, Pasquale Minervini, Philip H. S. Torr, Shimon Whiteson, Tim Rocktäschel

August 22, 2013