Seven local open VLMs were tested as a second reader, and Qwen3-VL 4B won
MaziyarPanahi · x · 2026-07-22
A local evaluation of seven small open VLMs found that second-reader behavior matters more than simple text reading.
- Qwen3-VL 4B got all 7 identifiers right in 6.9s with zero false alarms and was picked for the role.
- Gemma 4 E2B reached 6/7.
- Qwen3-VL 2B was fastest but only found 4/7.
- NVIDIA LocateAnything-3B drew perfect boxes but failed to judge what was actually personal.
- LFM2.5-VL 1.6B and InternVL3 2B read the patient name correctly, then incorrectly marked it as not personal.
The takeaway: detection is easy; judgment is the hard part. The poster also links an Apache 2.0 “Reader 1” and says both models are on Hugging Face.
Related event: Qwen3-VL Beats 6 Open-Source VLMs in Medical Image Desensitization Test(3 posts)→
More from Research
- Turing Motors says its CTO won gold in Kaggle’s 2026 ARC-AGI-linked contest — MeganRisdal · 2026-07-22
- Paper warns dubious Kaggle medical datasets are reaching both papers and clinics — EhudReiter · 2026-07-22
- Protein language models can learn homo-oligomer contacts from single sequences — anshulkundaje · 2026-07-22
- Commercial frontier models blocked attack forensics because they misread the responder — morqon · 2026-07-22
- LeCun’s JEPA pitch gets a concrete world-model paper behind it — nikola_mr64990 · 2026-07-22
- MIT opens a large free AI library with classic books, lecture notes, and courses — ZabihullahAtal · 2026-07-22