Principal Prototype Analysis on Manifold for Interpretable Reinforcement Learning

arXiv cs.LG / 3/31/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper addresses a growing interpretability gap in reinforcement learning as RL systems become more complex and harder to explain end-to-end reasoning patterns.
It builds on Prototype-Wrapper Networks (PW-Nets), but proposes automatically selecting the most suitable reference prototypes from data instead of requiring manually defined prototypes by domain experts.
The proposed method aims to preserve the tradeoff PW-Nets target—improving explainability while maintaining efficiency and strong RL performance.
Preliminary experiments on standard OpenAI Gym benchmarks show the approach achieves comparable performance to existing PW-Nets and remains competitive with black-box models.

Abstract

Recent years have witnessed the widespread adoption of reinforcement learning (RL), from solving real-time games to fine-tuning large language models using human preference data significantly improving alignment with user expectations. However, as model complexity grows exponentially, the interpretability of these systems becomes increasingly challenging. While numerous explainability methods have been developed for computer vision and natural language processing to elucidate both local and global reasoning patterns, their application to RL remains limited. Direct extensions of these methods often struggle to maintain the delicate balance between interpretability and performance within RL settings. Prototype-Wrapper Networks (PW-Nets) have recently shown promise in bridging this gap by enhancing explainability in RL domains without sacrificing the efficiency of the original black-box models. However, these methods typically require manually defined reference prototypes, which often necessitate expert domain knowledge. In this work, we propose a method that removes this dependency by automatically selecting optimal prototypes from the available data. Preliminary experiments on standard Gym environments demonstrate that our approach matches the performance of existing PW-Nets, while remaining competitive with the original black-box models.