VLM in a flash: I/O-Efficient Sparsification of Vision-Language Model via Neuron Chunking Kichang Yang†,1 · Seonjun Kim†,1 · Minjae Kim1 · Nairan Zhang2 · Chi Zhang3 · Youngki Lee1 † Equal Contribution 1 Seoul National University · 2 Meta · 3 Amazon Accepted to NeurIPS 2025 🚀 Code coming soon!