The Quantization Robustness of Diffusion Language Models in Coding Benchmarks