Residual Context Diffusion Language Models

Diffusion Large Language Models (dLLMs) have emerged as a promising alternative to purely autoregressive language models because they can decode multiple tokens in parallel. However, state-of-the-art block-wise dLLMs rely on a ``remasking" mechanism that decodes only the most confident tokens and discards the rest, effectively wasting computation. We demonstrate that recycling computation from the discarded tokens is beneficial, as these tokens retain contextual information useful for subsequent decoding iterations. In light of this, we propose Residual Context Diffusion (RCD), a module that converts these discarded token representations into contextual residuals and injects them back for the next denoising step. RCD uses a decoupled two-stage training pipeline to bypass the memory bottlenecks associated with backpropagation. We validate our method on both long CoT reasoning (SDAR) and short CoT instruction following (LLaDA) models. We demonstrate that a standard dLLM can be efficiently converted to the RCD paradigm with merely $\sim$300 million tokens. RCD consistently improves frontier dLLMs by 4--11 percentage points in accuracy with minimal extra computation overhead across a wide range of benchmarks. Notably, on the most challenging AIME tasks, RCD nearly doubles baseline accuracy and attains up to 4--5x fewer denoising steps at baseline's peak accuracy.

Diffusion Large Language Models (dLLMs) offer parallel decoding benefits but often trail autoregressive models in accuracy. This is partly due to the "remasking" strategy which discards intermediate computation on low-confidence tokens at each step. We propose Residual Context Diffusion (RCD), a novel paradigm that recycles this wasted computation into a contextual guiding signal. By treating the latent representations of undecoded tokens as residual updates, RCD progressively refines the model's knowledge across denoising steps. Our method achieves 4–11 percentage point accuracy gains across GSM8K and MATH500, more than doubles baseline accuracy on the challenging AIME24 benchmark, and requires up to 5× fewer denoising steps at equivalent accuracy levels.