ATLAS: Actor-Critic Task-Completion with Look-ahead Action Simulation
This addresses the need for more adaptable and efficient web-agents without requiring fine-tuning, though it is incremental as it builds on existing agent architectures.
The paper tackles the problem of web-agents failing to adapt to new environments without fine-tuning, by introducing ATLAS, which uses a cognitive map and action simulation to improve planning, achieving a 63% success rate on the WebArena-Lite Benchmark compared to 53.9% for previous state-of-the-art.
We observe that current state-of-the-art web-agents are unable to effectively adapt to new environments without neural network fine-tuning, without which they produce inefficient execution plans due to a lack of awareness of the structure and dynamics of the new environment. To address this limitation, we introduce ATLAS (Actor-Critic Task-completion with Look-ahead Action Simulation), a memory-augmented agent that is able to make plans grounded in a model of the environment by simulating the consequences of those actions in cognitive space. Our agent starts by building a "cognitive map" by performing a lightweight curiosity driven exploration of the environment. The planner proposes candidate actions; the simulator predicts their consequences in cognitive space; a critic analyzes the options to select the best roll-out and update the original plan; and a browser executor performs the chosen action. On the WebArena-Lite Benchmark, we achieve a 63% success rate compared to 53.9% success rate for the previously published state-of-the-art. Unlike previous systems, our modular architecture requires no website-specific LLM fine-tuning. Ablations show sizable drops without the world-model, hierarchical planner, and look-ahead-based replanner confirming their complementary roles within the design of our system