Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2606.00838 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910276496916480 |
|---|---|
| author | Subramanian, Vignesh Roy, Subhajit Bansal, Suguman |
| author_facet | Subramanian, Vignesh Roy, Subhajit Bansal, Suguman |
| contents | Inductive generalization is a framework for reinforcement learning (RL) generalization in which inductively related task instances admit inductively related policies. Prior work captures this structure via a higher-order policy-evolution function learned directly with RL, but suffers from poor training scalability: as training tasks grow, aggregated reward feedback becomes noisy and conflicting, destabilizing training and weakening generalization. We propose DIBS, a decoupled behavioral cloning approach that separates learning task-specific policies from learning the evolution function. We first learn individual teacher policies per task via standard RL, then fit the evolution function via behavioral cloning on teacher-labeled state-action pairs. This replaces noisy reward aggregation with dense, stable supervision. DIBS achieves significant improvements in both training stability and zero-shot generalization against existing RL and meta-RL algorithms. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2606_00838 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Decoupled Behavioral Cloning for Scalable Inductive Generalization in RL from Specifications Subramanian, Vignesh Roy, Subhajit Bansal, Suguman Artificial Intelligence Inductive generalization is a framework for reinforcement learning (RL) generalization in which inductively related task instances admit inductively related policies. Prior work captures this structure via a higher-order policy-evolution function learned directly with RL, but suffers from poor training scalability: as training tasks grow, aggregated reward feedback becomes noisy and conflicting, destabilizing training and weakening generalization. We propose DIBS, a decoupled behavioral cloning approach that separates learning task-specific policies from learning the evolution function. We first learn individual teacher policies per task via standard RL, then fit the evolution function via behavioral cloning on teacher-labeled state-action pairs. This replaces noisy reward aggregation with dense, stable supervision. DIBS achieves significant improvements in both training stability and zero-shot generalization against existing RL and meta-RL algorithms. |
| title | Decoupled Behavioral Cloning for Scalable Inductive Generalization in RL from Specifications |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2606.00838 |