Papers for

customer service bot teams

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Llm agents learn web tasks by building skills step by step

From Local Skills to Long-Horizon Web Tasks: Progressive Skill Exploration for LLM Web Agents

Abstract: Large language models (LLMs) are driving autonomous agents toward real-world environments, with the Web emerging as one of the most representative interaction scenarios. However, training Web agents with naive exploration remains challenging: successful trajectories are rare, and the sparse terminal rewards make it difficult for agents to obtain useful feedback from failed trajectories. Inspired by skill acquisition theory, we propose Skill-Composed Progressive Exploration (SCOPE), a short-to-long learning framework for long-horizon Web agents that decomposes complex task exploration into a gradual expansion process from local skills to complete task-solving capability. Specifically, SCOPE iteratively explores around local skills: Skill Discovery extracts reusable local skills and failure experiences from historical rollouts; Skill Orchestration composes discovered skills into high-level skill chains; and Skill Expansion leverages failure experiences aligned with the current skill chain to guide the agent toward exploring new subskills. Through this process, SCOPE progressively constructs task-completing skill chains across rollouts and internalizes them into the policy parameters through agentic on-policy distillation. Extensive experiments on WebArena, WebVoyager, and WebShop show that SCOPE consistently improves exploration dynamics and achieves 3.9% to 18.6% relative gains over the strongest baselines in each setting.

Mon 2 NovArtificial Intelligence

The gist

Using large language models (LLMs) to automate tasks on the web is hard because success happens rarely and feedback from failure is limited. The authors propose a method called SCOPE, which helps the agent gradually learn by first mastering small skills near its starting point, then combining these skills to solve bigger tasks over time. By learning from both successes and failures in this structured way, the agent improves its ability to complete complex web tasks. Experiments show this approach performs better than previous methods on web interaction benchmarks.

Open → 2611.03116v1