RepProbe Synthetic - MAE: leaderboard
RepProbe Synthetic - MAE. Generalist video-language models; counts 1–20, predictions clipped 0–40; Few 1–6, Mid 7–13, Many 14–20. Reported metric: MAE (lower is better). Primary published table, with source configuration boundaries preserved.
Metric: MAE (lower is better). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | InternVL3.5-8B | 6.5 | |
| 2 | LLaVA-OneVision-7B | 7.13 | |
| 3 | Cosmos-Reason2-8B | 7.75 | |
| 4 | Qwen3-VL-8B | 7.77 | |
| 5 | VideoChat-R1.5-7B | 7.96 | |
| 6 | Video-R1-7B | 8.37 | |
| 7 | LongVILA-R1-7B | 8.43 |
Interactive version: theaggregate.ai/benchmark?slug=repprobe-synthetic-mae · How It Works · Data refreshed daily, snapshot 2026-10-16.