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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Agents that can autonomously navigate the web through a graphical userinterface (GUI) using a unified action space (e.g., mouse and keyboard actions)can require very large amounts of domain-specific expert demonstrations toachieve good performance. Low sample efficiency is often exacerbated insparse-reward and large-action-space environments, such as a web GUI, whereonly a few actions are relevant in any given situation. In this work, weconsider the low-data regime, with limited or no access to expert behavior. Toenable sample-efficient learning, we explore the effect of constraining theaction space through $\textit{intent-based affordances}$ -- i.e., consideringin any situation only the subset of actions that achieve a desired outcome. Wepropose $\textbf{Code as Generative Affordances}$ $(\textbf{$\texttt{CoGA}$})$,a method that leverages pre-trained vision-language models (VLMs) to generatecode that determines affordable actions through implicit intent-completionfunctions and using a fully-automated program generation and verificationpipeline. These programs are then used in-the-loop of a reinforcement learningagent to return a set of affordances given a pixel observation. By greatlyreducing the number of actions that an agent must consider, we demonstrate on awide range of tasks in the MiniWob++ benchmark that: $\textbf{1)}$$\texttt{CoGA}$ is orders of magnitude more sample efficient than its RL agent,$\textbf{2)}$ $\texttt{CoGA}$'s programs can generalize within a family oftasks, and $\textbf{3)}$ $\texttt{CoGA}$ performs better or on par comparedwith behavior cloning when a small number of expert demonstrations isavailable.

Cracking the Code of Action: a Generative Approach to Affordances for Reinforcement Learning