Timezone: »
Spotlight
Goal Misgeneralization in Deep Reinforcement Learning
Lauro Langosco di Langosco · Jack Koch · Lee Sharkey · Jacob Pfau · David Krueger
We study \emph{goal misgeneralization}, a type of out-of-distribution robustness failure in reinforcement learning (RL). Goal misgeneralization occurs when an RL agent retains its capabilities out-of-distribution yet pursues the wrong goal. For instance, an agent might continue to competently avoid obstacles, but navigate to the wrong place. In contrast, previous works have typically focused on capability generalization failures, where an agent fails to do anything sensible at test time.We provide the first explicit empirical demonstrations of goal misgeneralization and present a partial characterization of its causes.
Author Information
Lauro Langosco di Langosco (University of Cambridge)
Jack Koch (Unaffiliated)
Lee Sharkey (University of Tuebingen)
Jacob Pfau (University of Edinburgh)
David Krueger (University of Cambridge)
Related Events (a corresponding poster, oral, or spotlight)
-
2022 Poster: Goal Misgeneralization in Deep Reinforcement Learning »
Wed. Jul 20th through Thu the 21st Room Hall E #926
More from the Same Authors
-
2021 : Objective Robustness in Deep Reinforcement Learning »
Lauro Langosco di Langosco · Lee Sharkey -
2021 Workshop: INNF+: Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models »
Chin-Wei Huang · David Krueger · Rianne Van den Berg · George Papamakarios · Ricky T. Q. Chen · Danilo J. Rezende -
2020 Workshop: INNF+: Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models »
Chin-Wei Huang · David Krueger · Rianne Van den Berg · George Papamakarios · Chris Cremer · Ricky T. Q. Chen · Danilo J. Rezende -
2019 Workshop: Invertible Neural Networks and Normalizing Flows »
Chin-Wei Huang · David Krueger · Rianne Van den Berg · George Papamakarios · Aidan Gomez · Chris Cremer · Aaron Courville · Ricky T. Q. Chen · Danilo J. Rezende -
2018 Poster: Neural Autoregressive Flows »
Chin-Wei Huang · David Krueger · Alexandre Lacoste · Aaron Courville -
2018 Oral: Neural Autoregressive Flows »
Chin-Wei Huang · David Krueger · Alexandre Lacoste · Aaron Courville