This benchmark resulted in significant accuracy drop upto 60% on a range of vision models, from the standard ResNet visual classifier to the latest foundation models like CLIP and MiniGPT‑4
I asked the community on 2024-07-15 05:45 UTC whether large language models can be used to troubleshoot or improve convolutional neural network (CNN) based vision models.
The post includes the exact question: "Can language models help me fix such issues in CNN based vision models?"