Self-Guided Plan Extraction for Instruction-Following Tasks with Goal-Conditional Reinforcement Learning

TL;DR

SuperIgor is a novel framework that uses self-guided plan extraction and iterative co-training of language models and RL agents to improve instruction-following in dynamic environments without manual dataset annotation.

Contribution

It introduces a self-learning mechanism for plan generation and refinement, reducing reliance on predefined subtasks and manual annotations.

Findings

SuperIgor agents follow instructions more strictly than baselines.

The framework demonstrates strong generalization to unseen instructions.

It effectively handles environments with rich dynamics and stochasticity.

Abstract

We introduce SuperIgor, a framework for instruction-following tasks. Unlike prior methods that rely on predefined subtasks, SuperIgor enables a language model to generate and refine high-level plans through a self-learning mechanism, reducing the need for manual dataset annotation. Our approach involves iterative co-training: an RL agent is trained to follow the generated plans, while the language model adapts and modifies these plans based on RL feedback and preferences. This creates a feedback loop where both the agent and the planner improve jointly. We validate our framework in environments with rich dynamics and stochasticity. Results show that SuperIgor agents adhere to instructions more strictly than baseline methods, while also demonstrating strong generalization to previously unseen instructions.