Yee et al.
New benchmark tests vision-language models under real-world visibility limits
SynDORBench evaluates how camera distance, lighting, and pixel density degrade multimodal perception, finding that model scale alone does not predict robustness.
Big Tech
Jeremy Stephen Gabriel Yee · Zhengkui Wang · Zhiyuan Zhang · Avinash Anand · Timothy Liu · Benedict Chan · +2 more
Singapore Institute of Technology · Singapore Management University · NVIDIA AI Technology Center
Research Digest··2 min read
The authors introduce SynDORBench, a benchmark of over 54,000 question-answer pairs that systematically degrade visibility according to the DORI standard (Detection, Observation, Recognition, Identification).
Why this paper
From NVIDIA AI Technology Center and 2 others
In one line
Pixel density and physical imaging constraints govern LVLM perceptual failure, not model scale alone.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§