Skip to content

Dataset · simular-ai

Sai agent run logs on OSWorld-V2

simular-ai/sai-osworld-v2-benchmark-runs

Text logs, trajectories, and evaluations of two complete runs of the Sai computer agent on the OSWorld-V2 benchmark.

What's inside

Contains detailed data on two complete runs (108 tasks each) of the Claude-Opus-5-based Sai computer agent in the OSWorld-V2 environment. Includes final scores, step-by-step trajectories, system and API logs, and run manifests in text format.

What you can do with it

  • Error and performance analysis of computer-use agents
  • Research on autonomous agent trajectories in operating system environments
Tasks
Other
Downloads this month
21.8K
Likes
2

How to load it

from datasets import load_dataset

ds = load_dataset("simular-ai/sai-osworld-v2-benchmark-runs")