WeDLM: Reconciling Diffusion Language Models with Standard Causal Attention for Fast Inference Paper • 2512.22737 • Published Dec 28, 2025 • 2