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

arXiv: https://arxiv.org/abs/2604.20601 · HTML: https://mlsanity.com/paper/2604.20601/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in RL training of language and agent models, one of 15 directions in Deep reinforcement learning methods.
- Builds on 5 of its 31 references that are in the map.
- First posted on arXiv 22 Apr 2026.

## Placed in

- Deep reinforcement learning methods / [RL training of language and agent models](https://mlsanity.com/deeprl/rl-training-of-language-and-agent-models/)

## Builds on (most cited first, 5 of 5)

- [BabyAI: A Platform to Study the Sample Efficiency of Grounded Language Learning](https://mlsanity.com/paper/1810.08272/) (2018)
- [MineDreamer: Learning to Follow Instructions via Chain-of-Imagination for Simulated-World Control](https://mlsanity.com/paper/2403.12037/) (2024)
- [SIMA 2: A Generalist Embodied Agent for Virtual Worlds](https://mlsanity.com/paper/2512.04797/) (2025)
- [CraftAssist: A Framework for Dialogue-enabled Interactive Agents](https://mlsanity.com/paper/1907.08584/) (2019)
- [Adapting Interleaved Encoders with PPO for Language-Guided Reinforcement Learning in BabyAI](https://mlsanity.com/paper/2510.23148/) (2025)

## Abstract (arXiv)

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.