Ultra-Fast Language Generation via Discrete Diffusion Divergence Instruct

• URL: http://arxiv.org/abs/2509.25035v2
• Date: Wed, 01 Oct 2025 17:45:09 GMT
• Title: Ultra-Fast Language Generation via Discrete Diffusion Divergence Instruct
• Authors: Haoyang Zheng, Xinyang Liu, Cindy Xiangrui Kong, Nan Jiang, Zheyuan Hu, Weijian Luo, Wei Deng, Guang Lin,
• Abstract summary: DiDi-Instruct is a training-based method that distills a few-step student for fast generation.<n>On OpenWebText, DiDi-Instruct achieves perplexity from 62.2 (8 NFEs) to 18.4 (128 NFEs)<n>These gains come with a negligible entropy loss (around $1%$) and reduce additional training wall-clock time by more than $20times$ compared to competing dLLM distillation methods.
• Score: 24.431216450821463
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Fast and high-quality language generation is the holy grail that people pursue in the age of AI. In this work, we introduce Discrete Diffusion Divergence Instruct (DiDi-Instruct), a training-based method that initializes from a pre-trained (masked) discrete diffusion language model (dLLM) and distills a few-step student for fast generation. The resulting DiDi-Instruct model achieves comparable or superior performance to its dLLM teacher and the GPT-2 baseline while enabling up to 64$\times$ acceleration. The theoretical foundation of DiDi-Instruct is a novel framework based on integral KL-divergence minimization, which yields a practical training algorithm. We further introduce grouped reward normalization, intermediate-state matching, and the reward-guided ancestral sampler that significantly improve training stability, model coverage, and inference quality. On OpenWebText, DiDi-Instruct achieves perplexity from 62.2 (8 NFEs) to 18.4 (128 NFEs), which outperforms prior accelerated dLLMs and GPT-2 baseline. These gains come with a negligible entropy loss (around $1\%$) and reduce additional training wall-clock time by more than $20\times$ compared to competing dLLM distillation methods. We further validate the robustness and effectiveness of DiDi-Instruct through extensive ablation studies, model scaling, and the generation of discrete protein sequences. In conclusion, DiDi-Instruct is an efficient yet effective distillation method, enabling language generation in the blink of an eye. We will release both code and models at github.com/haoyangzheng-ai/didi-instruct.

Related papers

• DiRL: An Efficient Post-Training Framework for Diffusion Language Models [54.405206032785706]
  Diffusion Language Models (dLLMs) have emerged as promising alternatives to Auto-Regressive (AR) models.<n>Existing methods suffer from computational inefficiency and objective mismatches between training and inference.<n>We introduce DiRL, an efficient post-training framework that tightly integrates FlexAttention-accelerated blockwise training with LMDeploy-optimized inference.
  arXiv  Detail & Related papers  (2025-12-23T08:33:19Z)
• TwinFlow: Realizing One-step Generation on Large Models with Self-adversarial Flows [25.487712175353035]
  We propose TwinFlow, a framework for training 1-step generative models.<n>Our method achieves a GenEval score of 0.83 in 1-NFE on text-to-image tasks.<n>Our approach matches the performance of the original 100-NFE model on GenEval and DPG-Bench benchmarks.
  arXiv  Detail & Related papers  (2025-12-03T07:45:46Z)
• Saber: An Efficient Sampling with Adaptive Acceleration and Backtracking Enhanced Remasking for Diffusion Language Model [98.35868970993232]
  Diffusion language models (DLMs) are emerging as a powerful and promising alternative to the dominant autoregressive paradigm.<n>We introduce efficient Sampling with Adaptive acceleration and Backtracking Enhanced Remasking (i.e., Saber) to achieve better inference speed and output quality in code generation.
  arXiv  Detail & Related papers  (2025-10-20T23:38:12Z)
• DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation [68.19756761027351]
  Diffusion large language models (dLLMs) are compelling alternatives to autoregressive (AR) models.<n>We investigate their denoising processes and reinforcement learning methods.<n>Our work provides deeper insight into the machinery of dLLM generation and offers an effective, diffusion-native RL training framework.
  arXiv  Detail & Related papers  (2025-06-25T17:35:47Z)
• Uni-Instruct: One-step Diffusion Model through Unified Diffusion Divergence Instruction [22.698086367085384]
  Uni-Instruct is motivated by our proposed diffusion expansion theory of the $f$-divergence family.<n>On the CIFAR10 generation benchmark, Uni-Instruct achieves record-breaking Frechet Inception Distance (FID) values of textbfemph1.46 for unconditional generation.<n>On the ImageNet-$64times 64$ generation benchmark, Uni-Instruct achieves a new SoTA one-step generation FID of textbfemph1.02, which outperforms its 79-step teacher diffusion with a significant improvement margin of 1.33.
  arXiv  Detail & Related papers  (2025-05-27T05:55:45Z)
• d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning [31.531278643184656]
  Recent large language models (LLMs) have demonstrated strong reasoning capabilities that benefits from online reinforcement learning (RL)<n>We propose d1, a framework to adapt pre-trained masked dLLMs into reasoning models via a combination of supervised finetuning (SFT) and RL.<n>We find that d1 yields the best performance and significantly improves performance of a state-of-the-art dLLM.
  arXiv  Detail & Related papers  (2025-04-16T16:08:45Z)
• David and Goliath: Small One-step Model Beats Large Diffusion with Score Post-training [8.352666876052616]
  We propose Diff-Instruct* (DI*), a data-efficient post-training approach for one-step text-to-image generative models.<n>Our method frames alignment as online reinforcement learning from human feedback.<n>Our 2.6B emphDI*-SDXL-1step model outperforms the 50-step 12B FLUX-dev model.
  arXiv  Detail & Related papers  (2024-10-28T10:26:19Z)
• Language Rectified Flow: Advancing Diffusion Language Generation with Probabilistic Flows [53.31856123113228]
  This paper proposes Language Rectified Flow (ours) Our method is based on the reformulation of the standard probabilistic flow models. Experiments and ablation studies demonstrate that our method can be general, effective, and beneficial for many NLP tasks.
  arXiv  Detail & Related papers  (2024-03-25T17:58:22Z)
• ACT-Diffusion: Efficient Adversarial Consistency Training for One-step Diffusion Models [59.90959789767886]
  We show that optimizing consistency training loss minimizes the Wasserstein distance between target and generated distributions. By incorporating a discriminator into the consistency training framework, our method achieves improved FID scores on CIFAR10 and ImageNet 64$times$64 and LSUN Cat 256$times$256 datasets.
  arXiv  Detail & Related papers  (2023-11-23T16:49:06Z)
• Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution [67.9215891673174]
  We propose score entropy as a novel loss that naturally extends score matching to discrete spaces. We test our Score Entropy Discrete Diffusion models on standard language modeling tasks.
  arXiv  Detail & Related papers  (2023-10-25T17:59:12Z)
• BOOT: Data-free Distillation of Denoising Diffusion Models with Bootstrapping [64.54271680071373]
  Diffusion models have demonstrated excellent potential for generating diverse images. Knowledge distillation has been recently proposed as a remedy that can reduce the number of inference steps to one or a few. We present a novel technique called BOOT, that overcomes limitations with an efficient data-free distillation algorithm.
  arXiv  Detail & Related papers  (2023-06-08T20:30:55Z)
• Adan: Adaptive Nesterov Momentum Algorithm for Faster Optimizing Deep Models [134.83964935755964]
  In deep learning, different kinds of deep networks typically need different extrapolations, which have to be chosen after multiple trials.<n>To relieve this issue and consistently improve the model training speed deep networks, we propose the ADAtive Nesterov momentum Transformer.
  arXiv  Detail & Related papers  (2022-08-13T16:04:39Z)
• Fast Sequence Generation with Multi-Agent Reinforcement Learning [40.75211414663022]
  Non-autoregressive decoding has been proposed in machine translation to speed up the inference time by generating all words in parallel. We propose a simple and efficient model for Non-Autoregressive sequence Generation (NAG) with a novel training paradigm: Counterfactuals-critical Multi-Agent Learning (CMAL) On MSCOCO image captioning benchmark, our NAG method achieves a performance comparable to state-of-the-art autoregressive models, while brings 13.9x decoding speedup.
  arXiv  Detail & Related papers  (2021-01-24T12:16:45Z)
• POINTER: Constrained Progressive Text Generation via Insertion-based Generative Pre-training [93.79766670391618]
  We present POINTER, a novel insertion-based approach for hard-constrained text generation. The proposed method operates by progressively inserting new tokens between existing tokens in a parallel manner. The resulting coarse-to-fine hierarchy makes the generation process intuitive and interpretable.
  arXiv  Detail & Related papers  (2020-05-01T18:11:54Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.