Explainable Reinforcement Learning via Interpretable Policy Distillation

Author(s)	Mr. Samyo Ranjan Jagdev, Mr. Achinta Kumar Palit, Ms. Subhashree Sibani Sahu
Country	India
Abstract	Abstract—Deep reinforcement learning achieves remarkable performance but often operates as a black box, limiting trust in safety-critical applications. Explainable Reinforcement Learning (XRL) aims to provide human-understandable explanations. We propose a unified framework for XRL via interpretable policy distillation, where a high-performing teacher policy is distilled into interpretable models such as decision trees or sparse linear policies. The distilled policy enables both global and local expla- nations while maintaining high fidelity. We provide theoretical guarantees on fidelity and performance, and perform extensive experiments with ablation studies. Results show that interpretable policies achieve competitive performance with high transparency, enabling deployment in real-world scenarios.
Keywords	Index Terms—Explainable Reinforcement Learning, Policy Distillation, Interpretability, Deep Reinforcement Learning
Field	Engineering
Published In	Volume 8, Issue 1, January-February 2026
Published On	2026-01-15
DOI	https://doi.org/10.36948/ijfmr.2026.v08i01.65614

About IJFMR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9687-828-838	editor@ijfmr.com

International Journal For Multidisciplinary Research