AliyunConsoleAgent: Training Web Agents in Real-World Cloud Environments via Distillation and Reinforcement Learning

2026-06-08

AI Summary

This work addresses the challenges of automating validation in cloud console environments, where inconsistencies between documentation and user interfaces result in extremely low manual test coverage (under 1%), and the high cost and deployment difficulty of large language models. To overcome these issues, we propose a two-stage training paradigm: first performing supervised fine-tuning (SFT) via distillation of trajectories from a state-of-the-art model, followed by policy optimization through reinforcement learning (GRPO) in real cloud environments. Our approach innovatively integrates distillation with reinforcement learning, featuring a high-certainty rollback mechanism and a rule-based evaluation protocol resistant to reward hacking. We further introduce a dual-channel reward model, Terraform-based resource provisioning, and rule-based rewards derived from audit logs. Evaluated on 278 tasks, our method achieves a 63.52% success rate—20.24 percentage points above the baseline—and closely matches the performance of the state-of-the-art model (within 1.82 pp) while reducing inference costs by 92%.

Abstract

We present AliyunConsoleAgent, a web agent framework for automated documentation verification in real-world cloud consoles. Major cloud platforms encompass hundreds of products with rapid feature iteration, causing console UIs to frequently diverge from their corresponding documentation. Verifying that documented procedures accurately reflect the current console and can be executed end-to-end demands an estimated 4 million recurring inspections annually, yet manual coverage remains below 1%. While agent systems built on frontier proprietary models achieve high success rates, their prohibitive cost and data privacy constraints preclude large-scale deployment. We propose a two-stage training paradigm: supervised fine-tuning (SFT) on distilled frontier-model trajectories, followed by reinforcement learning using Group Relative Policy Optimization (GRPO) and a dual-channel outcome reward model in real cloud environments. To support large-scale RL training, we construct a high-determinism rollout system featuring Terraform-based resource pre-provisioning and LLM-driven on-demand provisioning, which effectively isolates environment noise from the training signal. We further introduce a rule-based reward evaluation protocol grounded in backend audit logs, providing objective, reward-hacking-resistant outcome judgment. Our model evolves from mechanical instruction following to autonomous decision-making with cloud console and product-specific understanding. Experiments on a challenging 278-task benchmark where the best frontier model achieves only 65.34% demonstrate that AliyunConsoleAgent-32B achieves a 63.52% mean success rate -- a 20.24 percentage-point improvement over the base model, narrowing the gap to the best frontier proprietary model to 1.82 pp (bootstrap 95% CI [-1.27, 7.39]) -- at 92% lower inference cost.

Problem

• documentation verification
• cloud console
• web agents
• UI-documentation divergence
• automated testing

Innovation

• distillation
• reinforcement learning
• GRPO
• high-determinism rollout
• reward evaluation protocol