Reinforcement Learning of Implicit and Explicit Control Flow in Instructions

TL;DR

This paper introduces an attention-based reinforcement learning architecture that learns to flexibly follow complex instructions involving control flow, such as branching and looping, with zero-shot generalization to longer instructions in dynamic environments.

Contribution

The paper presents a novel attention-based architecture that learns control flow in instructions from reward signals, enabling flexible and generalizable instruction following in RL agents.

Findings

Achieves zero-shot generalization to longer instructions

Outperforms baseline recurrent architectures in instruction following tasks

Handles both explicit and implicit control flow in dynamic environments

Abstract

Learning to flexibly follow task instructions in dynamic environments poses interesting challenges for reinforcement learning agents. We focus here on the problem of learning control flow that deviates from a strict step-by-step execution of instructions -- that is, control flow that may skip forward over parts of the instructions or return backward to previously completed or skipped steps. Demand for such flexible control arises in two fundamental ways: explicitly when control is specified in the instructions themselves (such as conditional branching and looping) and implicitly when stochastic environment dynamics require re-completion of instructions whose effects have been perturbed, or opportunistic skipping of instructions whose effects are already present. We formulate an attention-based architecture that meets these challenges by learning, from task reward only, to flexibly attend to and condition behavior on an internal encoding of the instructions. We test the architecture's ability to learn both explicit and implicit control in two illustrative domains -- one inspired by Minecraft and the other by StarCraft -- and show that the architecture exhibits zero-shot generalization to novel instructions of length greater than those in a training set, at a performance level unmatched by two baseline recurrent architectures and one ablation architecture.