Date
AI Overview
Why it matters
Personalized learning path planning is often constrained by static profiles, while large language model approaches struggle to model sustained, long-term learning progression and align with evolving educational goals.
Study at a glance
- Question: Can an adaptive learning framework using reinforcement learning and learner state modeling improve personalized educational path planning?
- Setting and data: 22 undergraduate students in an online learning platform, alongside 300 real-world learning sessions with 14,584 interaction turns across 148 online courses.
- Method: Randomized controlled trial (controlled user study) and offline model evaluation.
- AI tools: Pxplore, GPT-4o, GPT-4o-mini, Qwen3-8B, Llama3.1-8B-Instruct, Claude-3.7-Sonnet
Key findings
- Undergraduate students using the Pxplore platform improved mean test scores from 61.81% to 90.09%, achieving a 28.28% knowledge gain compared to 14.18% in the control group.
- Pxplore trained with Group Relative Policy Optimization reached an average pedagogical alignment rate of 65.47%, outperforming fine-tuned models and prompting baselines.
- Students using Pxplore reported higher subjective satisfaction (4.64 out of 5) and motivation (4.73 out of 5) compared to a traditional retrieval-based baseline.
Summary drafted with AI from the full paper and may contain errors. See the full paper for details.
Personalized Learning Path Planning (PLPP) aims to design adaptive learning
paths that align with individual goals. While large language models (LLMs) show
potential in personalizing learning experiences, existing approaches often lack
mechanisms for goal-aligned planning. We introduce Pxplore, a novel framework
for PLPP that integrates a reinforcement-based training paradigm and an
LLM-driven educational architecture. We design a structured learner state model
and an automated reward function that transforms abstract objectives into
computable signals. We train the policy combining supervised fine-tuning (SFT)
and Group Relative Policy Optimization (GRPO), and deploy it within a
real-world learning platform. Extensive experiments validate Pxplore's
effectiveness in producing coherent, personalized, and goal-driven learning
paths. We release our code and dataset to facilitate future research.
Research Categories
What is the application?
Who is the user?
Who age?
Why use AI?
Publication Source
Publisher
arXiv
