Tag
DiffuSpace 发布 Clover-35B-A3B-Preview，首个开源块扩散语言模型，采用 35B 参数 MoE 骨干（约 3B 激活参数）和 128K 上下文窗口，宣称在发布时达到开源扩散语言模型的最先进性能。
Intern Lumina U2 is presented as the first diffusion language model using multi-codebook discrete visual tokens to unify image, video, and 3D generation with text in a single backbone, achieving leading performance and 4.45x faster image generation via stability-aware sampling.
This paper introduces block parallelism and context-sharded block parallelism (CSBP) to efficiently train long-context diffusion language models, achieving significant throughput improvements and better performance on benchmarks like SWE-bench Verified.
NVIDIA introduces TwoTower, a method that decouples context representation and denoising in diffusion language models, achieving 2.42x throughput while retaining 98.7% of autoregressive quality on a 30B MoE backbone.
NVIDIA Research introduces Nemotron-Labs-TwoTower, a diffusion language model that splits a 30B model into two halves for parallel token generation, achieving 2.42× faster generation while retaining 98.7% of original quality.
TAG-DLM unifies textual reasoning and graph message passing within a masked diffusion language model, enabling joint reasoning over text and graph topology for node classification and link prediction tasks.
The paper proposes Nemotron-TwoTower, a diffusion language model that decouples context representation and denoising using a frozen autoregressive tower and a trainable diffusion denoiser, achieving 98.7% of baseline quality with 2.42x throughput.
The author built Joey, a 170M parameter masked diffusion language model from scratch, trained on FineWeb-Edu and fine-tuned on DailyDialog, achieving fluent but incoherent sentences due to capacity limitations. The project highlights the differences from autoregressive LLMs and the lessons learned from building and debugging the system.
NVIDIA introduces Nemotron-Labs Diffusion, a family of diffusion language models that generate text in parallel and iteratively refine it, offering faster generation and the ability to revise previous tokens.
This solo-author ICML paper introduces Amortized Group Relative Policy Optimization (AGRPO) to enable effective reinforcement learning post-training for diffusion language models.
This research paper introduces Chainwash, a multi-step rewriting attack that effectively removes statistical watermarks from diffusion language model (LLaDA-8B-Instruct) outputs, reducing detection rates from 87.9% to 4.86% after five chained rewrites.
Introduces Token Time Continuous Diffusion (TTCD), a new diffusion language model that operates in continuous space with per-token times, outperforming discrete models at high speedups in conditional generation and Sudoku solving.