Papers
arxiv:2105.14111

Goal Misgeneralization in Deep Reinforcement Learning

Published on Jan 9, 2023
Authors:
,
,
,
,
,

Abstract

Reinforcement learning agents can retain skills out-of-distribution while pursuing incorrect goals, a failure mode distinct from capability breakdowns, with initial empirical examples and causal analysis provided.

We study goal misgeneralization, a type of out-of-distribution generalization failure in reinforcement learning (RL). Goal misgeneralization failures occur 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 formalize this distinction between capability and goal generalization, provide the first empirical demonstrations of goal misgeneralization, and present a partial characterization of its causes.

Community

Sign up or log in to comment

Get this paper in your agent:

hf papers read 2105.14111
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2105.14111 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2105.14111 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2105.14111 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.