Adapting Interleaved Encoders with PPO for Language-Guided Reinforcement Learning in BabyAI

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Mathur, Aryan, Ahmed, Asaduddin
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917045299314688
author Mathur, Aryan
Ahmed, Asaduddin
author_facet Mathur, Aryan
Ahmed, Asaduddin
contents Deep reinforcement learning agents often struggle when tasks require understanding both vision and language. Conventional architectures typically isolate perception (for example, CNN-based visual encoders) from decision-making (policy networks). This separation can be inefficient, since the policy's failures do not directly help the perception module learn what is important. To address this, we implement the Perception-Decision Interleaving Transformer (PDiT) architecture introduced by Mao et al. (2023), a model that alternates between perception and decision layers within a single transformer. This interleaving allows feedback from decision-making to refine perceptual features dynamically. In addition, we integrate a contrastive loss inspired by CLIP to align textual mission embeddings with visual scene features. We evaluate the PDiT encoders on the BabyAI GoToLocal environment and find that the approach achieves more stable rewards and stronger alignment compared to a standard PPO baseline. The results suggest that interleaved transformer encoders are a promising direction for developing more integrated autonomous agents.
format Preprint
id arxiv_https___arxiv_org_abs_2510_23148
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Adapting Interleaved Encoders with PPO for Language-Guided Reinforcement Learning in BabyAI
Mathur, Aryan
Ahmed, Asaduddin
Machine Learning
Artificial Intelligence
Image and Video Processing
I.2.6; I.2.9; I.5.4
Deep reinforcement learning agents often struggle when tasks require understanding both vision and language. Conventional architectures typically isolate perception (for example, CNN-based visual encoders) from decision-making (policy networks). This separation can be inefficient, since the policy's failures do not directly help the perception module learn what is important. To address this, we implement the Perception-Decision Interleaving Transformer (PDiT) architecture introduced by Mao et al. (2023), a model that alternates between perception and decision layers within a single transformer. This interleaving allows feedback from decision-making to refine perceptual features dynamically. In addition, we integrate a contrastive loss inspired by CLIP to align textual mission embeddings with visual scene features. We evaluate the PDiT encoders on the BabyAI GoToLocal environment and find that the approach achieves more stable rewards and stronger alignment compared to a standard PPO baseline. The results suggest that interleaved transformer encoders are a promising direction for developing more integrated autonomous agents.
title Adapting Interleaved Encoders with PPO for Language-Guided Reinforcement Learning in BabyAI
topic Machine Learning
Artificial Intelligence
Image and Video Processing
I.2.6; I.2.9; I.5.4
url https://arxiv.org/abs/2510.23148