OSWorld V2 Tasks: Python Task Classes for Operating System Benchmark

A gated dataset containing the official root-level Python task classes for OSWorld V2, created by xlangai. The task implementations are kept separate from public documentation to reduce benchmark leakage and prevent agents from finding answers online. The dataset was last updated on June 8, 2026.

Use Cases

• Benchmarking AI agents on operating system interaction tasks based on the described task classes.
• Developing task loaders and utilities for OSWorld V2 based on the provided Python code.
• Preventing evaluation leakage by using the gated task implementations as described.

Strengths

• Task implementations are gated to reduce benchmark leakage, as explicitly stated.
• Last updated date is provided: 2026-06-08 09:34:14.

Limitations

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

Provenance

Source

xlangai via Hugging Face.

Collection Method

Likely created as part of the OSWorld V2 benchmark project.

Freshness

Last updated 2026-06-08 09:34:14.