Researchers have developed RecVerse, a novel agent designed to more faithfully simulate human shopping behavior in e-commerce environments. This agent addresses two key challenges: memory limitations in long sessions and the optimization difficulties of current simulation methods. RecVerse employs a hierarchical memory system and is trained using trajectory-level reinforcement learning, enabling it to produce more realistic and intent-consistent user sessions. The team also released the User Simulation Benchmark (USB) dataset to facilitate further research in this area. AI
IMPACT This research could improve offline evaluation and RL training for e-commerce recommender systems, reducing the need for costly online A/B testing.
RANK_REASON The cluster describes a new research paper detailing a novel agent and dataset for simulating user behavior.
Read on Hugging Face Daily Papers →
- Alibaba Group
- National University of Singapore
- RecVerse
- Renmin University of China
- University of Chinese Academy of Sciences
- USB
- vision-language model
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
