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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Offline goal-conditioned RL (GCRL) offers a way to train general-purposeagents from fully offline datasets. In addition to being conservative withinthe dataset, the generalization ability to achieve unseen goals is anotherfundamental challenge for offline GCRL. However, to the best of our knowledge,this problem has not been well studied yet. In this paper, we studyout-of-distribution (OOD) generalization of offline GCRL both theoretically andempirically to identify factors that are important. In a number of experiments,we observe that weighted imitation learning enjoys better generalization thanpessimism-based offline RL method. Based on this insight, we derive a theoryfor OOD generalization, which characterizes several important design choices.We then propose a new offline GCRL method, Generalizable OfflinegoAl-condiTioned RL (GOAT), by combining the findings from our theoretical andempirical studies. On a new benchmark containing 9 independent identicallydistributed (IID) tasks and 17 OOD tasks, GOAT outperforms currentstate-of-the-art methods by a large margin.

What is Essential for Unseen Goal Generalization of Offline Goal-conditioned RL?