LASER: LLM Agent with State-Space Exploration for Web Navigation
This addresses the issue of sub-optimal performance in LLM-based web navigation for users needing robust interactive agents, though it is an incremental improvement over existing methods.
The paper tackled the problem of LLMs struggling with challenging scenarios like mistakes in web navigation due to forward-only execution, and proposed LASER, which models the task as state-space exploration to enable flexible backtracking. The result shows that LASER significantly outperforms previous methods and closes the gap with human performance on the WebShop task and amazon.com.
Large language models (LLMs) have been successfully adapted for interactive decision-making tasks like web navigation. While achieving decent performance, previous methods implicitly assume a forward-only execution mode for the model, where they only provide oracle trajectories as in-context examples to guide the model on how to reason in the environment. Consequently, the model could not handle more challenging scenarios not covered in the in-context examples, e.g., mistakes, leading to sub-optimal performance. To address this issue, we propose to model the interactive task as state space exploration, where the LLM agent transitions among a pre-defined set of states by performing actions to complete the task. This formulation enables flexible backtracking, allowing the model to recover from errors easily. We evaluate our proposed LLM Agent with State-Space ExploRation (LASER) on both the WebShop task and amazon.com. Experimental results show that LASER significantly outperforms previous methods and closes the gap with human performance on the web navigation task.