Research map / Large language models / Language, safety and interpretability
Accelerating Diffusion Large Language Models with SlowFast Sampling: The Three Golden Principles
Qingyan Wei, Yaojie Zhang, Zhiyuan Liu, Puyu Zeng, Yuxuan Wang, Biqing Qi, Dongrui Liu, Linfeng Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add SlowFast Sampling | LLaDA 8B · GSM8K | speedup: 1.00x -> 3.20x (score 69.83 -> 69.59) | Improved |
| add SlowFast Sampling instead of Fast-dLLM (Parallel) | Dream 7B · MMLU | score: 64.59 -> 75.13 | Improved |
| add SlowFast Sampling instead of Fast-dLLM (Parallel) | LLaDA 8B · BBH | speedup: 2.66x -> 5.24x | Improved |
| add SlowFast Sampling | LLaDA 8B · MMLU | score: 62.11 -> 66.56 | Improved |
| add SlowFast Sampling | Dream 7B · MMLU-pro | score: 24.14 -> 22.91 | Hurt |
| add SlowFast Sampling | Dream 7B · BBH | score: 51.83 -> 50.55 | Hurt |
| add dLLM-Cache to SlowFast Sampling | LLaDA 8B · GSM8K | speedup: 3.20x -> 5.93x (score 69.59 -> 69.60) | Improved |
| add dLLM-Cache to SlowFast Sampling (Kp=500, Kr=30) | LLaDA 8B · GPQA (8-shot, length 1024) | accuracy: 31.47 -> 28.79 with 34.22x speedup | Mixed |
| use SlowFast Sampling instead of semi-autoregressive sampling | LLaDA 8B · GSM8K | TPS and score: 5.44 -> 9.87 TPS; 66.41 -> 69.59 | Improved |
| vary stability-check hyperparameters Kmax, W_hist, sigma^2_stable (ablation) | SlowFast Sampling · GSM8K | accuracy and TPS | Mixed |
| raise tau_high_conf (ablation) | SlowFast Sampling · GSM8K | accuracy vs TPS | Mixed |
Closest papers by meaning
- Accelerating Diffusion LLMs via Adaptive Parallel Decoding (NeurIPS 2025 · similarity 0.94)
- Confidence-Based Decoding is Provably Efficient for Diffusion Language Models (NeurIPS 2026 · similarity 0.92)
- SimSD: Simple Speculative Decoding in Diffusion Language Models (NeurIPS 2026 · similarity 0.92)
- Fast-dLLM++: Fr\'{e}chet Profile Decoding for Faster Diffusion LLM Inference (NeurIPS 2026 · similarity 0.92)
- Fast and Fluent Diffusion Language Models via Convolutional Decoding and Rejective Fine-tuning (NeurIPS 2025 · similarity 0.92)
- Free Draft-and-Verification: Toward Lossless Parallel Decoding for Diffusion Large Language Models (NeurIPS 2026 · similarity 0.92)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
