# Vision-language models stay confident even when fog makes them blind

> A new arXiv study shows five small VLMs keep high confidence while accuracy drops to chance at around 12‑16 m visibility in fog.

Oossa · 2026-10-19 · https://oossa.com/en/vision-language-models-stay-confident-even-when-fog-makes-them-blind

Researchers led by Mochamad Rizal Fauzan posted a paper on arXiv on 2026‑10‑19. They tested five open‑source vision‑language models (2‑5 billion parameters) on 1,000 COCO images while gradually reducing visibility with simulated fog and water. Accuracy fell to chance when visibility was between 12.1 m and 15.9 m in fog, but the models still reported internal confidence scores of 0.68‑0.84 and verbal confidence up to 1.0. Only one model showed confidence that tracked accuracy.

## The facts

- Five VLMs, 2‑5 B parameters, tested on 1,000 COCO scenes
- Half‑way accuracy visibility (V_acc) in fog: 12.1‑15.9 m

## Why it matters

Users of VLMs in low‑visibility settings, like autonomous drones, may be misled by over‑confident outputs.

## Sources & references

1. [How Murky Before a VLM Knows It Cannot See? Accuracy, Confidence and Abstention on a Physical Visibility Axis](https://arxiv.org/abs/2610.22981) – arXiv cs.CV, 2026-10-19

Last updated: 2026-10-19