Quant Experts: Token-aware Adaptive Error Reconstruction with Mixture of Experts for Large Vision-Language Models Quantization

Quant Experts: Token-aware Adaptive Error Reconstruction with Mixture of Experts for Large Vision-Language Models Quantization explores Enable efficient Vision-Language Model deployment with adaptive token-aware quantization for reduced computational cost without sacrificing accuracy.. Commercial viability score: 6/10 in Model Optimization.