Early-Bird Decoding: Accelerating Diffusion LLMs with Learnable Block Sizes and Parallel Sampling

arXiv:2609.16450v1 Announce Type: cross Abstract: Diffusion large language models (dLLMs) offer a promising parallel decoding paradigm as an alternative to autoregressive generation through iterative unmasking. However, dLLMs typically require many steps before token confidence reaches the decoding threshold, resulting in inefficient inference even with block-wise KV caching. To accelerate dLLM inference, we for the first time propose an "early-bird (EB)" decoding framework, motivated by the observation that tokens with similarly low entropy tend to cluster and can be jointly decoded earlier, before reaching the confidence threshold. In particular, our EB-Decode framework integrates two key enablers: (1) a learnable network that adaptively groups tokens with similar uncertainty into variable-length blocks, rather than relying on fixed block sizes; (2) a position-aware sampler that learns to unmask tokens in parallel using fewer decoding steps within predicted variable-length blocks. Both components are developed without modifying pretrained dLLM weights and can therefore be directly deployed as plug-ins during serving, with negligible training and inference overhead. Extensive experiments across three models and four benchmarks consistently validate our observation and the effectiveness of EB-Decode, achieving 3.53-18.76imes higher throughput than the vanilla decoding method and up to 1.58imes higher throughput over the strongest baseline, Fast-dLLM, with comparable accuracy.

Related

• Dynamic-dLLM: Dynamic Cache-Budget and Adaptive Parallel Decoding for Training-Free Acceleration of Diffusion LLM
• Dependency-Aware Revocable Decoding for Efficient Diffusion Large Language Model Inference
• PRESTO: Prefix-Aligned Tree Drafting for Diffusion Speculative Decoding
• TAD: Temporal-Aware Trajectory Self-Distillation for Fast and Accurate Diffusion LLM

Source: arXiv cs.AI | 2026-09-16