Westworld: A Benchmark for Web Agents on Realistic Websites

Halluminate Westworld is a benchmark for evaluating web agents on tasks performed on realistic websites. The dataset was created by Halluminate and was last updated on Hugging Face on November 19, 2025. It includes a human-in-the-loop demo for manually running tasks.

Use Cases

• Benchmarking web agent performance based on tasks performed on realistic websites.
• Training reinforcement learning agents for web navigation based on the described benchmark tasks.
• Conducting human-in-the-loop evaluations using the provided demo interface.

Strengths

• Designed as a benchmark for a specific, emerging domain (web agents).
• Includes a human-in-the-loop demo for manual task execution, which suggests practical validation.
• Last updated on November 19, 2025, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The description metadata is limited; actual data quality and structure require manual inspection after download.

Provenance

Source

Halluminate

Collection Method

Likely constructed for benchmarking web agent tasks on simulated or real websites.

Freshness

Last updated 2025-11-19 06:03:38.