: Visual Understanding of VLMs in Really Dense Scenes

Paul Gavrikov1,4, Wei Lin2, M. Jehanzeb Mirza3, Soumya Jahagirdar4, Muhammad Huzaifa4, Sivan Doveh5, Serena Yeung-Levy5, James Glass3, Hilde Kuehne4,6
1 Independent Researcher 2 JKU Linz 3 MIT CSAIL 4 Tübingen AI Center 5 Standford 6 MIT-IBM Watson AI Lab
CVPR, 2026
🤗 Dataset 🚀 Evaluation Server Paper Code arXiv

🔥 News

  • [2026/04/23] Added results for Qwen 3.6-27B (non-thinking).
  • [2026/04/16] Added results for Qwen 3.6-35B-A3B (non-thinking).
  • [2026/04/14] Added results for Gemma 4, InternVL 3.5, and Qwen 3.5 (all non-thinking).
  • [2026/02/21] VisualOverload is accepted into CVPR 2026!

Abstract

Is basic visual understanding really solved in state-of-the-art VLMs? We present VisualOverload, a slightly different visual question answering (VQA) benchmark comprising 2,720 question–answer pairs, with privately held ground-truth responses. Unlike prior VQA datasets that typically focus on near global image understanding, VisualOverload challenges models to perform simple, knowledge-free vision tasks in densely populated (or, overloaded) scenes. Our dataset consists of high-resolution scans of public-domain paintings that are populated with multiple figures, actions, and unfolding subplots set against elaborately detailed backdrops. We manually annotated these images with questions across six task categories to probe for a thorough understanding of the scene. We hypothesize that current benchmarks overestimate the performance of VLMs, and encoding and reasoning over details is still a challenging task for them, especially if they are confronted with densely populated scenes. Indeed, we observe that even the best model (o3) out of 37 tested models only achieves 19.6% accuracy on our hardest test split and overall 69.5% accuracy on all questions. Beyond a thorough evaluation, we complement our benchmark with an error analysis that reveals multiple failure modes, including a lack of counting skills, failure in OCR, and striking logical inconsistencies under complex tasks. Altogether, VisualOverload exposes a critical gap in current vision models and offers a crucial resource for the community to develop better models.

  • 2,720 question-answer pairs
  • 6 tasks and 3 levels of difficulty
  • fresh image data
  • all images are public domain

Leaderboard

You can benchmark your own model by submitting your predictions to our evaluation server. If you want your submission to appear on the public leaderboard, please follow the instructions to open a GitHub issue. Note: all results reported below were obtained without any reasoning/thinking modes (unless explictly noted as reasoning-mode) and typically using greedy decoding.
Model Special Inference Activity Attributes Counting OCR Reasoning Scene Easy Medium Hard Total

Example Questions

BibTeX

@InProceedings{Gavrikov_2026_visualoverload,
      author = {Paul Gavrikov and Wei Lin and M. Jehanzeb Mirza and Soumya Jahagirdar and Muhammad Huzaifa and Sivan Doveh and Serena Yeung-Levy and James Glass and Hilde Kuehne},
      title = {{VisualOverload}: Probing Visual Understanding of VLMs in Really Dense Scenes}, 
      booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
      month = {June},
      year = {2026}
}