Step-by-step trajectory viewers for WebOperator runs (gpt-4o), browsable on GitHub Pages. Screenshots are loaded from the dataset on Hugging Face.
All run data (search trees, logs, trajectories, evaluations, configs, screenshots and analysis scripts) is on Hugging Face: https://huggingface.co/datasets/mahirlabibdihan/weboperator-experiments
| Benchmark | Viewers |
|---|---|
| WebArena (812 tasks) | webarena/main/ |
| WebArena-lite ablation study (155 tasks, 10 runs) | webarena-lite/ablations/ |
Viewer URL pattern:
https://mahirlabibdihan.github.io/WebOperator-Experiments/<benchmark>/tasks/task_<id>/trajectory.html