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 are interested in the autonomous acquisition of repertoires of skills.Language-conditioned reinforcement learning (LC-RL) approaches are great toolsin this quest, as they allow to express abstract goals as sets of constraintson the states. However, most LC-RL agents are not autonomous and cannot learnwithout external instructions and feedback. Besides, their direct languagecondition cannot account for the goal-directed behavior of pre-verbal infantsand strongly limits the expression of behavioral diversity for a given languageinput. To resolve these issues, we propose a new conceptual approach tolanguage-conditioned RL: the Language-Goal-Behavior architecture (LGB). LGBdecouples skill learning and language grounding via an intermediate semanticrepresentation of the world. To showcase the properties of LGB, we present aspecific implementation called DECSTR. DECSTR is an intrinsically motivatedlearning agent endowed with an innate semantic representation describingspatial relations between physical objects. In a first stage (G -> B), itfreely explores its environment and targets self-generated semanticconfigurations. In a second stage (L -> G), it trains a language-conditionedgoal generator to generate semantic goals that match the constraints expressedin language-based inputs. We showcase the additional properties of LGB w.r.t.both an end-to-end LC-RL approach and a similar approach leveragingnon-semantic, continuous intermediate representations. Intermediate semanticrepresentations help satisfy language commands in a diversity of ways, enablestrategy switching after a failure and facilitate language grounding.

Grounding Language to Autonomously-Acquired Skills via Goal Generation