Search papers, labs, and topics across Lattice.
This paper identifies spatial and temporal redundancy as key bottlenecks in the decoding process of Diffusion Large Language Models (dLLMs). To address this, they propose $R^2$-dLLM, a framework that reduces redundancy both during inference, via confidence aggregation and temporal stabilization, and during training, via redundancy-aware fine-tuning. Experiments show that $R^2$-dLLM achieves up to a 75% reduction in decoding steps while maintaining competitive generation quality.
Ditch the slow lane: $R^2$-dLLM turbocharges diffusion language models by slashing decoding steps by up to 75% without sacrificing quality.
Diffusion Large Language Models (dLLMs) have emerged as a promising alternative to autoregressive generation by enabling parallel token prediction. However, practical dLLM decoding still suffers from high inference latency, which limits deployment. In this work, we observe that a substantial part of this inefficiency comes from recurring redundancy in the decoding process, including spatial redundancy caused by confidence clusters and positional ambiguity, and temporal redundancy caused by repeatedly remasking predictions that have already stabilized. Motivated by these patterns, we propose $R^2$-dLLM, a unified framework for reducing decoding redundancy from both inference and training perspectives. At inference time, we introduce training-free decoding rules that aggregate local confidence and token predictions, and finalize temporally stable tokens to avoid redundant decoding steps. We further propose a redundancy-aware supervised fine-tuning pipeline that aligns the model with efficient decoding trajectories and reduces reliance on manually tuned thresholds. Experiments demonstrate that $R^2$-dLLM consistently reduces the number of decoding steps by up to 75% compared to existing decoding strategies, while maintaining competitive generation quality across different models and tasks. These results validate that decoding redundancy is a central bottleneck in dLLMs, and that explicitly reducing it yields substantial practical efficiency gains.
