[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-nvidia-nemotron-diffusion-ar-dllm-tri-modal":3,"news-related-2b5b7c66-7289-45db-b5b7-dea67882310c":39},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":26,"news_slug":32,"published_at":33,"created_at":34,"modified_at":35,"is_published":36,"publish_type":37,"image_url":13,"view_count":38},"2b5b7c66-7289-45db-b5b7-dea67882310c","NVIDIA 发布 Nemotron-Labs Diffusion：三模态语言模型统一 AR 与扩散解码","传统的语言模型生成文本采用自回归（AR）方式，即逐个 token 生成，每个新 token 都依赖之前的所有 token。这种方式虽然稳定可靠，但存在明显的性能瓶颈：每次生成都需要完整的模型前向传播，且 GPU 大部分时间消耗在内存读写而非计算上。同时，AR 模型一旦生成 token 就不容修改，导致错误会不断累积传播。\n\nNVIDIA 于 5 月 23 日在 Hugging Face 博客上发布了 Nemotron-Labs Diffusion 语言模型，带来一种全新的解决思路。该模型是首个三模态语言模型，能够在同一架构内统一自回归、扩散和自推测解码（self-speculation decoding）三种生成模式。通过联合 AR-diffusion 目标训练，模型可以根据部署场景和并发需求动态切换生成模式，在精度和吞吐量之间实现灵活权衡。\n\n在 NVIDIA H100 GPU 上的测试显示，扩散模式相比传统自回归解码在吞吐量上有显著提升，同时保持了可比的精度水平。这种并行多 token 生成加迭代精化的方法，还有一个额外优势：模型能够在生成过程中修订已输出的 token，而 AR 模型一旦 token 生成即成定局无法回退。这一特性使扩散语言模型在需要fill-in-the-middle和文本修订的场景中更具优势。\n\n从工程角度看，Nemotron-Labs Diffusion 的最大价值在于提供了一种内置的性能调优手段。开发者可以通过减少精化步数来降低推理预算，在延迟敏感场景下换取更高吞吐量，而无需更换模型或引入辅助蒸馏模型。这一特性对需要在大批量、低延迟与高精度之间反复权衡的生产环境尤为重要。\n\n扩散语言模型并非要取代自回归范式，而是提供了一种正交的能力维度。两者各有适用场景，Nemotron-Labs Diffusion 允许模型在运行时根据实际情况选择最优模式，而非在训练时就锁定一种路径。这或许是未来高性能推理引擎的一种标准架构方向。","https:\u002F\u002Fhuggingface.co\u002Fblog\u002Fnvidia\u002Fnemotron-labs-diffusion","474eef8c-e0c3-46cf-adee-c089558220f9",[10,14,17,20,23],{"id":11,"name":12,"slug":12,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":21,"name":22,"slug":22,"description":13,"color":13},"8dac812d-3839-4abe-a855-5f56ec9515fd","nvidia",{"id":24,"name":25,"slug":25,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[27],{"id":28,"lang":29,"title":30,"summary":31,"content":13},"eba70c97-e0af-4b4f-9f38-809886194ad1","en","Nemotron-Labs Diffusion unifies AR and diffusion decoding","Traditional language models generate text autoregressively (AR) — token by token, each new token depending on all prior tokens. While stable and reliable, this approach has clear performance bottlenecks: every generation needs a full model forward pass, and the GPU spends most of its time on memory I\u002FO rather than computation. Also, AR models cannot revise a token once it's generated, so errors accumulate and propagate.\n\nOn May 23, NVIDIA published the Nemotron-Labs Diffusion language model on the Hugging Face blog — a fresh take on the problem. The model is the first tri-modal language model that unifies three generation modes within one architecture: autoregressive, diffusion, and self-speculation decoding. Trained with a joint AR-diffusion objective, the model can dynamically switch generation modes based on deployment scenario and concurrency needs, achieving flexible trade-offs between accuracy and throughput.\n\nTesting on NVIDIA H100 GPUs shows diffusion mode offers significant throughput gains over traditional autoregressive decoding while maintaining comparable accuracy. This parallel multi-token generation plus iterative refinement approach has an additional advantage: the model can revise already-output tokens during generation, whereas an AR model cannot walk back a token once it's committed. This makes diffusion language models stronger in scenarios requiring fill-in-the-middle and text revision.\n\nFrom an engineering perspective, Nemotron-Labs Diffusion's biggest value is providing a built-in performance-tuning lever. Developers can reduce inference budget by cutting refinement steps, trading higher throughput for latency-sensitive scenarios, without swapping models or introducing auxiliary distilled models. This is especially important in production environments that repeatedly juggle high-volume, low-latency, and high-precision needs.\n\nDiffusion language models are not about replacing the autoregressive paradigm, but adding an orthogonal capability axis. Each has its applicable scenarios; Nemotron-Labs Diffusion lets the model pick the optimal mode at runtime based on actual conditions, rather than locking the path at training time. This may well be a standard architectural direction for future high-performance inference engines.","nvidia-nemotron-diffusion-ar-dllm-tri-modal","2026-05-23T04:10:00Z","2026-05-23T04:06:47.580169Z","2026-08-19T02:08:40.142862Z",true,"agent",178,{"items":40},[41,46,51,56,61,66],{"id":42,"title":43,"news_slug":44,"published_at":45},"cac485ab-a429-4ebc-88c6-a1f924f978ff","AWS 开源 KeysAndValues:微调时就让模型学会“遗忘”,单张 A100 撑住 128K","aws-keysvalues-sparse-attention-finetuning","2026-08-26T05:20:00+00:00",{"id":47,"title":48,"news_slug":49,"published_at":50},"c94bdf86-5de9-49fe-8c98-0f5c47611bfe","SGLang v0.5.18 发布:大模型冷启动提速 2.38 倍,710 个 PR 都改了什么","sglang-v0-5-18-cold-start-2-38x","2026-08-24T23:15:00+00:00",{"id":52,"title":53,"news_slug":54,"published_at":55},"ce70384a-990b-4994-bfb6-27775be45661","TensorRT Edge-LLM 0.10.0：边端第一个统一的 C++ 多模态推理栈","tensorrt-edge-llm-0-10-multimodal-runtime","2026-08-23T00:00:00+00:00",{"id":57,"title":58,"news_slug":59,"published_at":60},"92433e6b-113a-4ada-af77-fbb8995a9850","LFM2.5-DSpark 开源:300M 草稿模型让端侧推理快 2.87 倍,输出零损耗","lfm2-5-dspark-draft-models","2026-08-21T21:10:00+00:00",{"id":62,"title":63,"news_slug":64,"published_at":65},"a8b9d045-0f4c-4596-baa7-060955365877","TensorRT Edge-LLM 0.10.0：Qwen3.8-27B Day-0 上车，边缘 LLM 推理再加速","tensorrt-edge-llm-qwen3-8-27b-day0","2026-08-21T15:00:00+00:00",{"id":67,"title":68,"news_slug":69,"published_at":70},"926f89fc-5ed5-4170-bfac-931d3a31b6a4","腾讯混元开源 AngelSpec 投机解码框架：DFly 在 Hy3-A21B 上取得 1.98–2.40× 加速","tencent-angelspec-spec-decoding-hy3-dfly","2026-07-30T00:00:00+00:00"]