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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

With strong capabilities of reasoning and a generic understanding of theworld, Large Language Models (LLMs) have shown great potential in buildingversatile embodied decision making agents capable of performing diverse tasks.However, when deployed to unfamiliar environments, we show that LLM agents facechallenges in efficiently gathering necessary information, leading tosuboptimal performance. On the other hand, in unfamiliar scenarios, humanindividuals often seek additional information from their peers before takingaction, leveraging external knowledge to avoid unnecessary trial and error.Building upon this intuition, we propose \textit{Asking Before Action} (ABA), amethod that empowers the agent to proactively query external sources forpertinent information using natural language during their interactions in theenvironment. In this way, the agent is able to enhance its efficiency andperformance by mitigating wasteful steps and circumventing the difficultiesassociated with exploration in unfamiliar environments. We empirically evaluateour method on an embodied decision making benchmark, ALFWorld, and demonstratethat despite modest modifications in prompts, our method exceeds baseline LLMagents by more than $40$%. Further experiments on two variants of ALFWorldillustrate that by imitation learning, ABA effectively retains and reusesqueried and known information in subsequent tasks, mitigating the need forrepetitive inquiries. Both qualitative and quantitative results exhibitremarkable performance on tasks that previous methods struggle to solve.

Asking Before Action: Gather Information in Embodied Decision Making with Language Models