AgentTrek: Agent Trajectory Synthesis via Guiding Replay with Web Tutorials
What it is
AgentTrek is a framework for synthesizing high-quality agent trajectories by guiding the replay of web tutorials. It uses these trajectories to train browser-use agents that can perform general web tasks autonomously.
The problem it solves
The high cost and difficulty of collecting high-quality, diverse trajectories for training autonomous web agents.
What you could use it for
- Training autonomous browser agents for web navigation
- Synthesizing training data for LLM-based agents
- Evaluating agent performance on benchmarks like WebArena and MiniWoB
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.