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

Offline goal-conditioned reinforcement learning (GCRL) promisesgeneral-purpose skill learning in the form of reaching diverse goals frompurely offline datasets. We propose $\textbf{Go}$al-conditioned$f$-$\textbf{A}$dvantage $\textbf{R}$egression (GoFAR), a novelregression-based offline GCRL algorithm derived from a state-occupancy matchingperspective; the key intuition is that the goal-reaching task can be formulatedas a state-occupancy matching problem between a dynamics-abiding imitator agentand an expert agent that directly teleports to the goal. In contrast to priorapproaches, GoFAR does not require any hindsight relabeling and enjoysuninterleaved optimization for its value and policy networks. These distinctfeatures confer GoFAR with much better offline performance and stability aswell as statistical performance guarantee that is unattainable for priormethods. Furthermore, we demonstrate that GoFAR's training objectives can bere-purposed to learn an agent-independent goal-conditioned planner from purelyoffline source-domain data, which enables zero-shot transfer to new targetdomains. Through extensive experiments, we validate GoFAR's effectiveness invarious problem settings and tasks, significantly outperforming priorstate-of-art. Notably, on a real robotic dexterous manipulation task, while noother method makes meaningful progress, GoFAR acquires complex manipulationbehavior that successfully accomplishes diverse goals.

How Far I'll Go: Offline Goal-Conditioned Reinforcement Learning via $f$-Advantage Regression