Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2604.20601 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915949286785024 |
|---|---|
| author | Volovikova, Zoya Sorokin, Nikita Lukashevskiy, Dmitriy Panov, Aleksandr Skrynnik, Alexey |
| author_facet | Volovikova, Zoya Sorokin, Nikita Lukashevskiy, Dmitriy Panov, Aleksandr Skrynnik, Alexey |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_20601 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Self-Guided Plan Extraction for Instruction-Following Tasks with Goal-Conditional Reinforcement Learning Volovikova, Zoya Sorokin, Nikita Lukashevskiy, Dmitriy Panov, Aleksandr Skrynnik, Alexey Artificial Intelligence Computation and Language 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. |
| title | Self-Guided Plan Extraction for Instruction-Following Tasks with Goal-Conditional Reinforcement Learning |
| topic | Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2604.20601 |