EgoExo-Next: leaderboard
EgoExo-Next (2026) tests whether vision-language models can reason about how an observed action trajectory will look next: 2,503 human-curated four-choice questions from six public egocentric and ego-exo video sources, where each option is a candidate visual state (an image) rather than a text label. Four subtasks cover egocentric next-state prediction, ego-to-exo and exo-to-ego state correspondence, exocentric next-state prediction and their composition (Ego-to-Exo Next-State), each split into indoor and outdoor scenes. The board is the paper's Avg., the unweighted mean accuracy over the eight subtask-by-scene settings, zero-shot, mean of three runs; chance is 25% and human accuracy is 98.55%.
Metric: Accuracy (%), unweighted mean over the eight subtask-by-scene settings (four-choice visual-option questions, chance 25%), zero-shot, mean of three runs. Source: arxiv.org. Saturation forecast: Around 2029. 9 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 43.81 | #102 |
| 2 | GPT-5.2 | 43.16 | #190 |
| 3 | Qwen 3 VL 235B A22B | 40.93 | #486 |
| 4 | Qwen 3 VL 32B | 37.03 | #624 |
| 5 | InternVL3.5-38B | 36.81 | #1026 |
| 6 | Qwen 3 VL 8B | 36.27 | #927 |
| 7 | Claude Opus 4.6 | 33.59 | #114 |
| 8 | InternVL3.5-8B | 32.93 | #1618 |
| 9 | LLaVA-OneVision-1.5-8B | 31.85 | #1876 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=egoexo-next · How It Works · Data refreshed daily, snapshot 2026-11-22.