TVWorld-N: leaderboard
Remote-control TV navigation (TVWorld-N, Ma et al., 2026): on offline graphs of five Google TV interface areas (configuration, display, audio, apps and channels), an agent sees screenshots with a focus highlight and presses remote keys (arrows, OK, exit, home and settings) to move from a start page to a goal page given as text or as a screenshot, within 50 steps. The score is the success rate over 500 tasks, in percent; higher is better.
Metric: Success Rate (%; 500 tasks: 5 Google TV graphs x 50 start-goal pairs x text and image goals). Source: arxiv.org. Saturation forecast: Around February 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 66.4 | #159 |
| 2 | GPT-5 Mini | 60.2 | #349 |
| 3 | Qwen 3 VL 32B Instruct | 39 | #582 |
| 4 | Claude Haiku 4.5 | 25.4 | #568 |
| 5 | Qwen 3 VL 8B Instruct | 20 | #982 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=tvworld-n · How It Works · Data refreshed daily, snapshot 2026-11-12.