Abstract
RecVerse is a GUI-grounded agent that uses hierarchical memory and trajectory-level reinforcement learning to simulate realistic multi-turn e-commerce shopping sessions.
Simulating realistic user shopping behavior underpins offline evaluation and reinforcement learning in e-commerce scenarios. While recent LLM- and VLM-based simulators have made encouraging progress, reproducing a real browsing session remains difficult for two reasons. (i) Memory Challenge: a shopping session spans dozens of pages, yet existing agents either discard long-range observation histories, losing the evolving user state, or naively concatenate them, overwhelming the context window and even degrading simulation quality. (ii) Optimization Challenge: current user simulators are typically supervised to match each logged action via imitation or step-level rewards; the resulting sessions often display unrealistic patterns, such as over-exploration or excessive passivity, which per-step supervision can neither detect nor correct. To address the above challenges, we present RecVerse, a GUI-grounded simulation agent that perceives pages through screenshots and produces faithful multi-turn trajectories. For the memory challenge, RecVerse adopts a cognitive-inspired hierarchical memory: Working Memory for short-term focus, Episodic Memory for in-session traces, and Preference Memory for high-level intent, with memory updates treated as actions so that the agent adaptively learns when and what to memorize. For the optimization challenge, RecVerse is optimized with a trajectory-level RL objective that scores entire sessions, aligning both macro-level action-type distributions and micro-level shopping intent with real users. We further release USB (User Simulation Benchmark), an interactive e-commerce GUI trajectory dataset for multi-turn user simulation. Experiments show that RecVerse significantly outperforms existing baselines in both behavioral fidelity and intent consistency.
Community
✨ RecVerse: Towards Faithful Simulation of Human Shopping Behavior 🛒
A GUI-grounded shopping agent that interacts with interfaces through screenshots 🖥️ like a real user, retains relevant information using a cognitively inspired hierarchical memory, and is trained end-to-end with trajectory-level reinforcement learning, producing simulated sessions that faithfully reflect how real people browse and shop online.
📄 Paper: Towards Faithful Simulation of Human Shopping Behavior (arXiv:2608.20707)
🏛️ Affiliations: Renmin University of China · UCAS · NUS · Alibaba Group
🤔 Why should you care?
Faithful user simulators power 🛍️ offline evaluation, counterfactual analysis, and RL recommender training — without burning online A/B traffic.
But current simulators still stumble on:
📚 Memory blow‑up across long sessions
🎯 Step‑wise imitation that reproduces noise and yields over‑active or overly passive behavior
💡RecVerse's Two Big Ideas
🧠 Cognitive‑Inspired Hierarchical Memory
- 👁️ Working Memory recent screenshots + mindset (FIFO)
- 📝 Episodic Memory session‑level textual event trace
- ❤️ Preference Memory distilled long‑term user intent
🪄 Memory‑as‑Action — the agent learns when and what to remember via RL.
🏹 Trajectory‑Aligned RL
- 📊 Macro reward aligns action‑type distributions
- 🔬 Micro reward uses a 3‑tier category tree for dense intent signals
- ✅ Format reward keeps outputs valid
🗂️ USB Benchmark
The first interactive GUI benchmark supporting multi‑turn agentic RL.
- 🖼️ 5,274 trajectories
- 🎬 69,842 actions
- 🏷️ 90,095 items (41/517/2,256 categories)
- 👤 5,222 users
🏆 Results
- vs. STA (best GUI baseline): F1 +68%, HR +77%, HCO +41% 📈
- Head‑to‑head human eval: RecVerse beats STA in 92% of cases 🥇
- Falls within the high-fidelity behavioral band, while others remain in the over-active or over-passive zone.
Get this paper in your agent:
hf papers read 2608.20707 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper
Collections including this paper 0
No Collection including this paper