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

The rise of generalist large-scale models in natural language and vision hasmade us expect that a massive data-driven approach could achieve broadergeneralization in other domains such as continuous control. In this work, weexplore a method for learning a single policy that manipulates various forms ofagents to solve various tasks by distilling a large amount of proficientbehavioral data. In order to align input-output (IO) interface among multipletasks and diverse agent morphologies while preserving essential 3D geometricrelations, we introduce morphology-task graph, which treats observations,actions and goals/task in a unified graph representation. We also developMxT-Bench for fast large-scale behavior generation, which supports proceduralgeneration of diverse morphology-task combinations with a minimal blueprint andhardware-accelerated simulator. Through efficient representation andarchitecture selection on MxT-Bench, we find out that a morphology-task graphrepresentation coupled with Transformer architecture improves the multi-taskperformances compared to other baselines including recent discretetokenization, and provides better prior knowledge for zero-shot transfer orsample efficiency in downstream multi-task imitation learning. Our worksuggests large diverse offline datasets, unified IO representation, and policyrepresentation and architecture selection through supervised learning form apromising approach for studying and advancing morphology-task generalization.

A System for Morphology-Task Generalization via Unified Representation and Behavior Distillation