Multi-task reinforcement learning (RL) aims to simultaneously learn policies for solving many tasks. Several prior works have found that relabeling past experience with different reward functions can improve sample efficiency. Relabeling methods typically pose the question: if, in hindsight, we assume that our experience was optimal for some task, for what task was it optimal?

Multi-task reinforcement learning (RL) aims to simultaneously learn policies for solving many tasks. Several prior works have found that relabeling past experience with different reward functions can improve sample efficiency. Relabeling methods typically pose the question: if, in hindsight, we assume that our experience was optimal for some task, for what task was it optimal? In this paper we show that inverse RL is a principled mechanism for reusing experience across tasks. We use this idea to generalize goal-relabeling techniques from prior work to arbitrary types of reward functions.

Multi-task reinforcement learning (RL) aims to simultaneously learn policies for solving many tasks. Several prior works have found that relabeling past experience with different reward functions can improve sample efficiency. Relabeling methods typically pose the question: if, in hindsight, we assume that our experience was optimal for some task, for what task was it optimal? In this paper we show that inverse RL is a principled mechanism for reusing experience across tasks. We use this idea to generalize goal-relabeling techniques from prior work to arbitrary types of reward functions.

History is rife with fakes. In 1983, the German magazine Stern announced that it had acquired previously undocumented diaries written by Hitler, a find British historian Hugh Trevor-Roper initially heralded as "an archive of great historical significance." In reality, however, an illustrator named Konrad Kujau had penned the volumes himself. Thanks to the scrutiny of historians at the German Federal Archive, they were soon revealed to be forgeries and the so-called "Hitler Diaries" became a cautionary tale about media frenzies. Imagine, however, if experts couldn't readily identify the diaries as fraudulent.