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 investigate the challenge of task planning for multi-task embodied agentsin open-world environments. Two main difficulties are identified: 1) executingplans in an open-world environment (e.g., Minecraft) necessitates accurate andmulti-step reasoning due to the long-term nature of tasks, and 2) as vanillaplanners do not consider how easy the current agent can achieve a givensub-task when ordering parallel sub-goals within a complicated plan, theresulting plan could be inefficient or even infeasible. To this end, we propose"$\underline{D}$escribe, $\underline{E}$xplain, $\underline{P}$lan and$\underline{S}$elect" ($\textbf{DEPS}$), an interactive planning approach basedon Large Language Models (LLMs). DEPS facilitates better error correction oninitial LLM-generated $\textit{plan}$ by integrating $\textit{description}$ ofthe plan execution process and providing self-$\textit{explanation}$ offeedback when encountering failures during the extended planning phases.Furthermore, it includes a goal $\textit{selector}$, which is a trainablemodule that ranks parallel candidate sub-goals based on the estimated steps ofcompletion, consequently refining the initial plan. Our experiments mark themilestone of the first zero-shot multi-task agent that can robustly accomplish70+ Minecraft tasks and nearly double the overall performances. Further testingreveals our method's general effectiveness in popularly adopted non-open-endeddomains as well (i.e., ALFWorld and tabletop manipulation). The ablation andexploratory studies detail how our design beats the counterparts and provide apromising update on the $\texttt{ObtainDiamond}$ grand challenge with ourapproach. The code is released at https://github.com/CraftJarvis/MC-Planner.

Describe, Explain, Plan and Select: Interactive Planning with Large Language Models Enables Open-World Multi-Task Agents