jewon lee

This report describes our approach to the Efficient Qwen Competition, where the goal is to enable low-latency serving of Qwen3.5-4B on a resource-constrained NVIDIA A10G GPU. Our system combines a quantized target model with speculative decoding. To recover accuracy, we apply quantization-aware distillation to the target model while retaining the original quantization grid.

Conflict of Interests The author declares no conflicts of interest. Supporting Information Filename Description cctc70499-sup-0001-SuppMat.docx7.9 MB Supporting Information File 1: cctc70499-sup-0001-SuppMat. docx. References 1, , , , and , “Plasma Technology Applications in Proton Exchange Membrane Fuel Cell Systems,” Chemcatchem 16 (2024): e202301372.