HAVEN - Saliency Ranking (With Document): leaderboard
Saliency ranking on HAVEN (2026), a video understanding benchmark with aligned frames, shots, descriptions and full video documents, with the full video document added to the shot frames. The score is the Spearman rank correlation with the annotated ranking, from -1 to 1; higher is better.
Metric: Shot saliency ranking: Spearman rank correlation (-1 to 1) between the predicted and the true order by importance, frames plus the full video document, zero-shot on a random sample of 100 videos per source dataset; higher is better. Source: arxiv.org. Saturation forecast: Around April 2027. 6 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 8B Instruct | 0.218 | #1043 |
| 2 | GPT-5.2 | 0.19 | #190 |
| 3 | InternVideo2.5-Chat-8B | 0.178 | #2049 |
| 4 | VideoLLaMA3-7B | 0.164 | #2064 |
| 5 | GLM-4.6V-Flash | 0.136 | #1049 |
| 6 | Qwen 2.5 VL 7B Instruct | 0.07 | #1919 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=haven-saliency-ranking-with-document · How It Works · Data refreshed daily, snapshot 2026-11-22.