AgentHijack - Subtitle: leaderboard
OSWorld Ubuntu tasks in AgentHijack (Sun et al., 2026) with a large music-player subtitle overlaid at the bottom of the screen. Percent of tasks completed, judged by OSWorld's execution-based checkers; higher is better.
Metric: Task Success Rate (%; subtitle overlay). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | UI-TARS-1.5-7B | 22.75 | |
| 2 | UI-TARS-72B-DPO | 19.32 | |
| 3 | UI-TARS-7B-DPO | 15.59 |
Interactive version: theaggregate.ai/benchmark?slug=agenthijack-subtitle · How It Works · Data refreshed daily, snapshot 2026-11-12.