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

Recent advancements in reasoning abilities of Large Language Models (LLM) haspromoted their usage in problems that require high-level planning for robotsand artificial agents. However, current techniques that utilize LLMs for suchplanning tasks make certain key assumptions such as, access to datasets thatpermit finetuning, meticulously engineered prompts that only provide relevantand essential information to the LLM, and most importantly, a deterministicapproach to allow execution of the LLM responses either in the form of existingpolicies or plan operators. In this work, we propose LgTS (LLM-guidedTeacher-Student learning), a novel approach that explores the planningabilities of LLMs to provide a graphical representation of the sub-goals to areinforcement learning (RL) agent that does not have access to the transitiondynamics of the environment. The RL agent uses Teacher-Student learningalgorithm to learn a set of successful policies for reaching the goal statefrom the start state while simultaneously minimizing the number ofenvironmental interactions. Unlike previous methods that utilize LLMs, ourapproach does not assume access to a propreitary or a fine-tuned LLM, nor doesit require pre-trained policies that achieve the sub-goals proposed by the LLM.Through experiments on a gridworld based DoorKey domain and a search-and-rescueinspired domain, we show that generating a graphical structure of sub-goalshelps in learning policies for the LLM proposed sub-goals and theTeacher-Student learning algorithm minimizes the number of environmentinteractions when the transition dynamics are unknown.

LgTS: Dynamic Task Sampling using LLM-generated sub-goals for Reinforcement Learning Agents