English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Building sample-efficient agents that generalize out-of-distribution (OOD) inreal-world settings remains a fundamental unsolved problem on the path towardsachieving higher-level cognition. One particularly promising approach is tobegin with low-dimensional, pretrained representations of our world, whichshould facilitate efficient downstream learning and generalization. By training240 representations and over 10,000 reinforcement learning (RL) policies on asimulated robotic setup, we evaluate to what extent different properties ofpretrained VAE-based representations affect the OOD generalization ofdownstream agents. We observe that many agents are surprisingly robust torealistic distribution shifts, including the challenging sim-to-real case. Inaddition, we find that the generalization performance of a simple downstreamproxy task reliably predicts the generalization performance of our RL agentsunder a wide range of OOD settings. Such proxy tasks can thus be used to selectpretrained representations that will lead to agents that generalize.

The Role of Pretrained Representations for the OOD Generalization of Reinforcement Learning Agents