Diffusion Models

41 items

RESEARCHarXiv CS.CL·26d ago

Differences in Text Generated by Diffusion and Autoregressive Language Models

This research explores the intrinsic differences in text generated by Diffusion Language Models (DLMs) and Autoregressive Language Models (ARMs), finding that DLMs show lower n-gram entropy but higher semantic coherence and diversity. Controlled experiments reveal that DLM training objectives enhance coherence and diversity through bidirectional context, while decoding algorithms are responsible for entropy reduction.

Diffusion Models language models NLP text generation

ARTICLEDEV.to AI·5/11/2026

Before the image knows what it is

The text explores the brief moment during image generation by diffusion models, when noise organizes into intention before taking shape. This moment of latency is where the art resides, before the image defines itself.

Diffusion Models creativity AI art Generative AI

NEWSDEV.to AI·18d ago

6.4 Claim Puts Nemotron-Labs Diffusion in AI Fast Lane

NVIDIA's Nemotron-Labs Diffusion aims to accelerate AI applications by tackling the one-token bottleneck through parallel generation of multiple tokens. This new diffusion language model claims up to 6.4 times higher tokens per forward pass, significantly benefiting latency-sensitive AI products like coding assistants and agent workflows.

Diffusion Models language models AI NVIDIA

RESEARCHHugging Face Blog·18d ago

Towards Speed-of-Light Text Generation with Nemotron-Labs Diffusion Language Models

This content discusses the development of Nemotron-Labs' diffusion language models aimed at achieving exceptionally high-speed text generation. The focus is on technical advancements to optimize the rapidity of text production.

Diffusion Models language models Nemotron-Labs text generation

NEWSDEV.to AI·9d ago

Bonsai Image 4B: difusión de 1 bit que corre en un iPhone

PrismML launched Bonsai Image 4B, a family of image generation models using 1-bit or ternary weights to run high-quality diffusion on local devices like iPhones. This innovation results in an 8.3x model compression, reducing it from 7.75 GB to 0.93 GB, while retaining up to 95% of the original quality.

Diffusion Models Edge AI image generation PrismML

RESEARCHDEV.to AI·20d ago

DualFashion: Dual-Diffusion Transformer Generates Outfit Images & Text

DualFashion is a dual-diffusion Transformer architecture that jointly generates fashion item images and textual descriptions. It outperforms state-of-the-art methods on iFashion and Polyvore-U benchmarks for generative outfit recommendation, providing interpretable outputs.

Diffusion Models image generation text generation fashion AI

RESEARCHDEV.to AI·29d ago

AI/ML Research Digest — May 09, 2026

This AI/ML research digest covers advancements in latent diffusion models for multimodal generation, focusing on efficiency and extending capabilities from images to video. It also highlights innovations in modular expert routing for neural networks and adaptive compute methods to optimize sequential decision-making processes.

Diffusion Models multimodal AI LLM Agents machine learning

RESEARCHarXiv CS.LG·4/9/2026

$S^3$: Stratified Scaling Search for Test-Time in Diffusion Language Models

O trabalho propõe $S^3$ (Stratified Scaling Search), um método de busca guiado por verificador para melhorar a qualidade de geração em modelos de linguagem de difusão durante o tempo de inferência. Ele realoca a computação no processo de denoising, avaliando e reamostrando seletivamente candidatos promissores para favorecer saídas de maior qualidade.

Diffusion Models search algorithms language models inference

RESEARCHarXiv CS.LG·5/1/2026

Simple Self-Conditioning Adaptation for Masked Diffusion Models

Masked diffusion models (MDMs) discard clean-state predictions for tokens that remain masked, limiting cross-step refinement. This paper proposes Self-Conditioned Masked Diffusion Models (SCMDM), a post-training adaptation that conditions each denoising step on the model's own previous clean-state predictions. This enhances performance without significant architectural changes or extra evaluations.

Diffusion Models model adaptation deep learning machine learning

RESEARCHarXiv CS.CL·5/8/2026

Chainwash: Multi-Step Rewriting Attacks on Diffusion Language Model Watermarks

This paper investigates multi-step rewriting attacks on diffusion language model watermarks, which are used to verify AI text authorship. The findings show that watermarked texts can have their detection compromised after multiple rewrites by other language models, even those unaware of the watermark key.

Diffusion Models language models AI watermarking security

RESEARCHarXiv CS.CL·19d ago

FlowLM: Few-Step Language Modeling via Diffusion-to-Flow Adaptation

FlowLM introduces a novel flow matching language model, adapted from pre-trained diffusion models through efficient fine-tuning. This method enables high-quality, few-step text generation that significantly outperforms traditional diffusion sampling with fewer training epochs.

Diffusion Models language models machine learning text generation

RESEARCHarXiv CS.LG·25d ago

Beyond Mode-Seeking RL: Trajectory-Balance Post-Training for Diffusion Language Models

This paper introduces TraFL, a novel post-training approach for diffusion language models that addresses "trajectory locking" observed in reward-maximizing methods. TraFL, a trajectory-balance objective, outperforms other methods across mathematical reasoning and code generation benchmarks.

Diffusion Models language models reinforcement learning machine learning

RESEARCHarXiv CS.LG·29d ago

Conditional generation of antibody sequences with classifier-guided germline-absorbing discrete diffusion

This research introduces a novel approach for conditional generation of antibody sequences, addressing limitations in current protein language models by better modeling somatic variation and enabling flexible classifier-guided generation. It proposes discrete diffusion fine-tuning and germline absorbing diffusion for improved antibody design.

Antibody Design Diffusion Models computational biology protein language models

RESEARCHarXiv CS.LG·27d ago

TMPO: Trajectory Matching Policy Optimization for Diverse and Efficient Diffusion Alignment

Trajectory Matching Policy Optimization (TMPO) addresses reward hacking in reinforcement learning for diffusion models, which often causes mode collapse and degrades generative diversity. It replaces scalar reward maximization with trajectory-level reward distribution matching, using a Softmax Trajectory Balance objective to align policy probabilities with a reward-induced Boltzmann distribution.

Diffusion Models reinforcement learning AI alignment Generative AI

RESEARCHarXiv CS.LG·27d ago

LEAP: Unlocking dLLM Parallelism via Lookahead Early-Convergence Token Detection

Diffusion Language Models (dLLMs) face scalability limits in parallelism due to overly conservative confidence thresholds that hinder their potential for highly parallel processing. This paper introduces LEAP, a training-free, plug-and-play method that improves dLLM parallelism by detecting early-converging tokens, thereby accelerating decoding.

Diffusion Models Parallel Computing AI large language models

RESEARCHarXiv CS.AI·11d ago

Orthogonal Concept Erasure for Diffusion Models

This research paper investigates the limitations of current concept erasure methods for mitigating undesired content in diffusion models. It identifies that additive parameter updates in editing-based methods cause entanglement between concept semantics and overall generative capacity, proposing a new solution to enhance precision and preservation.

Diffusion Models machine learning Concept Erasure AI safety

RESEARCHarXiv CS.LG·6d ago

Geometry-Aware Tabular Diffusion

Geometry-Aware Tabular Diffusion (GATD) is introduced to improve tabular synthesis by augmenting denoisers with pairwise angles and lengths computed from column value differences. It achieves state-of-the-art performance with fewer parameters, reducing Shape and Trend error, and showing that explicit relational supervision drives the gains.

Diffusion Models data synthesis deep learning machine learning

RESEARCHarXiv CS.CL·12d ago

ICG: Improving Cover Image Generation via MLLM-based Prompting and Personalized Preference Alignment

The paper proposes ICG, a novel framework for personalized cover image generation that integrates MLLM-based prompting with preference alignment. It utilizes semantic features and user embeddings to contextualize the diffusion model and adopts a multi-reward learning strategy to address the lack of labeled supervision.

personalization Diffusion Models MLLMs image generation

RESEARCHDEV.to AI·5/4/2026

Learning to Efficiently Sample from Diffusion Probabilistic Models

This research focuses on developing more efficient methods for sampling from Diffusion Probabilistic Models, aiming to reduce the computational cost and time associated with generating high-quality samples. It explores novel algorithms to accelerate the sampling process while maintaining the fidelity of the generated data.

Diffusion Models generative models machine learning Sampling Efficiency

RESEARCHYannic Kilcher (YouTube)·12/27/2025

TiDAR: Think in Diffusion, Talk in Autoregression (Paper Analysis)

This content provides an analysis of a research paper exploring the TiDAR model. The model integrates concepts of diffusion and autoregression for processing.

Diffusion Models AI models Paper analysis Machine learning research

TiDAR: Think in Diffusion, Talk in Autoregression (Paper Analysis)