EvoGUI-Bench - Successor Discrimination: leaderboard
The successor probe of EvoGUI-Bench (Tsinghua University, 2026): pick the logged next screen over a distractor from another trajectory, the same domain or a later step of the same workflow. Accuracy in percent; higher is better.
Metric: Pairwise successor accuracy (%; share of 1,000 items in which the model picks the logged next screen over a cross-trajectory, same-domain or later-step distractor; on EvoGUI-Bench, 3,000 zero-shot visual question answering probes mined from Mind2Web and WebLINX web trajectories across 120 domains, screenshots at max_image_side 512, temperature 0; higher is better). Source: arxiv.org. Saturation forecast: Around February 2027. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 83.6 | #144 |
| 2 | Doubao-Seed-2.0-Pro-260215 | 80.9 | #152 |
| 3 | Kimi K2.6 | 79.4 | #178 |
| 4 | Qwen 3.5 122B A10B | 77.8 | #331 |
| 5 | GLM-5V Turbo | 76.9 | #344 |
| 6 | Kimi K2.5 | 76.3 | #256 |
| 7 | MiMo-V2.5 | 76.1 | #354 |
| 8 | GPT-5.4 | 75.4 | #123 |
| 9 | Qwen 3.5 35B A3B | 74.7 | #482 |
| 10 | Qwen 3.6 35B A3B | 74.7 | #462 |
| 11 | Qwen 3.5 27B | 74.1 | #359 |
| 12 | GPT-5.4 Nano | 73.5 | #635 |
| 13 | Qwen 3.6 27B | 72.6 | #341 |
| 14 | Grok 4.20 | 72.3 | #297 |
| 15 | Qwen 3.5 2B | 71.3 | #4810 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=evogui-bench-successor-discrimination · How It Works · Data refreshed daily, snapshot 2026-11-13.