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

We study the problem of learning goal-conditioned policies in Minecraft, apopular, widely accessible yet challenging open-ended environment fordeveloping human-level multi-task agents. We first identify two main challengesof learning such policies: 1) the indistinguishability of tasks from the statedistribution, due to the vast scene diversity, and 2) the non-stationary natureof environment dynamics caused by partial observability. To tackle the firstchallenge, we propose Goal-Sensitive Backbone (GSB) for the policy to encouragethe emergence of goal-relevant visual state representations. To tackle thesecond challenge, the policy is further fueled by an adaptive horizonprediction module that helps alleviate the learning uncertainty brought by thenon-stationary dynamics. Experiments on 20 Minecraft tasks show that our methodsignificantly outperforms the best baseline so far; in many of them, we doublethe performance. Our ablation and exploratory studies then explain how ourapproach beat the counterparts and also unveil the surprising bonus ofzero-shot generalization to new scenes (biomes). We hope our agent could helpshed some light on learning goal-conditioned, multi-task agents in challenging,open-ended environments like Minecraft.

Open-World Multi-Task Control Through Goal-Aware Representation Learning and Adaptive Horizon Prediction