#Diffusionmodels - Twitter Hashtag

5 days ago

High-order ODE solvers stabilize zero-shot music editing without retraining. FlowSonic preserves harmonic & structural fidelity while enabling semantic modifications. #MusicGeneration #DiffusionModels https://t.co/3lgGr85gol

0

7

Furong Huang

@furongh

19 days ago

ICML 1/7: Scheduling Thoughts — Learning the Order of Thought in Diffusion Language Models. https://t.co/JYaGXYwjoU #ICML2026 #DiffusionModels #LLMReasoning

1

3

1

0

533

David @LiDavid2002

20 days ago

#ICML #MachineLearning #ArtificialIntelligence #AI #DeepLearning #GenerativeAI #DiffusionModels #LanguageModels #ComputerVision #ImageSuperResolution #AIResearch #Research #Seoul #Korea

0

2

0

72

Soumick @SoumickAI

26 days ago

7/8 Everything is open. Code, trained model weights, summary statistics and polygenic scores: 🔗 https://t.co/baRYH7EsV0 📄 Paper: https://t.co/iWp6exHAft #MachineLearning #GWAS #PRS #CardiacMRI #UKBiobank #DiffusionModels #ML #AI #ComputerVision

1

2

1

0

397

Rishubh Parihar @RishubhParihar

about 1 month ago

8/8 🙏 Huge thanks to my amazing co-authors: Ayush Raina, @rvbabuiisc, and @OPatashnik! Check out our preprint and examples on the project page! 🔗 https://t.co/Tohy8MkwE5 Let me know your thoughts! 👇 #DiffusionModels #ComputerVision #MachineLearning #CVPR #AIresearch

0

2

0

238

Amirhossein Kazerouni @amirkz99

about 2 months ago

9/ Project page: https://t.co/ThOKAV6PDp Paper: https://t.co/ERrQXYjdNv #ComputerVision #ImageRestoration #DiffusionModels #FaceRestoration #CVPR2026

0

3

0

86

Liang Zheng

@LiangZheng_06

about 2 months ago

Going to #CVPR2026. Great to catch up with old friends and meet new friends. Talk 1: diffusion post-training in AIMS workshop https://t.co/VWDdNj7hVv Talk 2: REPA-E and NanoGen in EDGE workshop https://t.co/Orud7ESSZX DM for ☕️ chat. #CVPR2026 #GenerativeAI #DiffusionModels

0

9

1

2

519

David @LiDavid2002

2 months ago

📄 Paper: https://t.co/ATpyG6rYwn 💻 Code: https://t.co/Vad4bnNXjN 📢 Invitation post: https://t.co/1Ny00y2nHN Many thanks to the organizers for the invitation, @jdeschena, @ssahoo_, @zhihanyang_! 🙌 #ICML2026 #DiffusionModels #LanguageModels #GenerativeAI #MachineLearning

Discrete Diffusion Reading Group

@diffusion_llms

2 months ago

📢 May 18 (Mon): IDLM: Inverse-distilled Diffusion Language Models 🤔Diffusion Language Models (DLMs) have recently achieved strong results in text generation. However, their multi-step sampling leads to slow inference, limiting practical use. 💡To address this, the authors extend Inverse Distillation, a technique originally developed to accelerate continuous diffusion models, to the discrete setting. However, this extension introduces both theoretical and practical challenges. 🔧To overcome these challenges, the authors first provide a theoretical result demonstrating that their inverse formulation admits a unique solution, thereby ensuring valid optimization. They then introduce gradient-stable relaxations to support effective training. 📊As a result, experiments on multiple DLMs show that their method, Inverse-distilled Diffusion Language Models (IDLM), reduces the number of inference steps by 4×—64×, while preserving the teacher model’s entropy and generative perplexity. This Monday, David Li (https://t.co/8geY1UUQ72) and Nikita Gushchin (https://t.co/L4nc4Gz7ZC) will present their jointly led paper, which was recently accepted at ICML 2026. Collaborators of this work include: Dmitry Abulkhanov (@dabulkhanov_), Eric Moulines (https://t.co/tK9ctSzw0r), Ivan Oseledets (@oseledetsivan), Maxim Panov (@maxim_panov), Alexander Korotin (https://t.co/RTSk7HAc3v) Paper link: https://t.co/L8MniD5GBu

diffusion_llms's tweet photo. 📢 May 18 (Mon): IDLM: Inverse-distilled Diffusion Language Models

🤔Diffusion Language Models (DLMs) have recently achieved strong results in text generation. However, their multi-step sampling leads to slow inference, limiting practical use.

💡To address this, the authors extend Inverse Distillation, a technique originally developed to accelerate continuous diffusion models, to the discrete setting. However, this extension introduces both theoretical and practical challenges.

🔧To overcome these challenges, the authors first provide a theoretical result demonstrating that their inverse formulation admits a unique solution, thereby ensuring valid optimization. They then introduce gradient-stable relaxations to support effective training.

📊As a result, experiments on multiple DLMs show that their method, Inverse-distilled Diffusion Language Models (IDLM), reduces the number of inference steps by 4×—64×, while preserving the teacher model’s entropy and generative perplexity.

This Monday, David Li (https://t.co/8geY1UUQ72) and Nikita Gushchin (https://t.co/L4nc4Gz7ZC) will present their jointly led paper, which was recently accepted at ICML 2026.

Collaborators of this work include: Dmitry Abulkhanov (@dabulkhanov_), Eric Moulines (https://t.co/tK9ctSzw0r), Ivan Oseledets (@oseledetsivan), Maxim Panov (@maxim_panov), Alexander Korotin (https://t.co/RTSk7HAc3v)

Paper link: https://t.co/L8MniD5GBu

2

36

14

10

10K

0

10

1

464

Onkar @_Onkara

3 months ago

Our paper “PODiff” has been accepted at ICML 2026. PODiff introduces diffusion in Proper Orthogonal Decomposition space for efficient and uncertainty aware scientific super resolution. Preprint: https://t.co/w6FcWg83xJ #ICML2026 #MachineLearning #DiffusionModels #ScientificML

1

0

77

MLV Lab @MLVLab

3 months ago

We are presenting our work, “SPRINT: Sparse-Dense Residual Fusion for Efficient Diffusion Transformers” at ICLR 2026 in Brazil! Time: Sat, Apr 25, 2026 • 3:15 PM – 5:45 PM Location: Pavilion 4 P4-3017 Paper: https://t.co/rvKkwodSal #ICLR2026 #Efficient #DiT #DiffusionModels

MLVLab's tweet photo. We are presenting our work,
“SPRINT: Sparse-Dense Residual Fusion for Efficient Diffusion Transformers” at ICLR 2026 in Brazil!

Time: Sat, Apr 25, 2026 • 3:15 PM – 5:45 PM
Location: Pavilion 4 P4-3017

Paper: https://t.co/rvKkwodSal

#ICLR2026 #Efficient #DiT #DiffusionModels https://t.co/2lSL7waIFG

0

3

1

0

158

Ant Open Source

@ant_oss

3 months ago

👋 Introducing LLaDA2.0-Uni — the 💥first unified MoE multimodal model in the LLaDA2.0 series of @TheInclusionAI, designed for native multimodal understanding and generation. #dLLM #inclusionAI #LLaDA #Multimodal #Opensource #DiffusionModels 1/2 Key features: 🧠Chain-of-Thought Generation LLaDA2.0-Uni doesn't just generate images — it thinks first. Through reasoning-augmented training, the model performs step-by-step reasoning before visual generation. Achieves 0.78 on WISE-Bench with thinking mode.

3

20

2

1

937

Weian Mao @WeianMaoX

3 months ago

Long video generation is bottlenecked by KV cache explosion. TriAttention v0.2.0 cuts LongLive's KV cache by 50% — no quality loss, plug-and-play for AR video diffusion. Would love PRs and benchmarks on more AR video models. #VideoGeneration #DiffusionModels #LongLive

Yukang Chen

@yukangchen_

3 months ago

We are excited to announce the release of TriAttention-v0.2.0!🚀 Compared with the initial version, the key updates include: 1⃣ Wider engine: Now supports both SGLang and vLLM⚙️ 2⃣Broader hardware: Adds support for NVIDIA Spark DGX, Apple Silicon, and AMD GPUs 💻 3⃣Video generation🎥 Beyond LLM / OpenClaw, TriAttention effectively compresses KV cache during AR diffusion inference — for example, reducing 50% KV cache for LongLive with no performance degradation. Check it out: https://t.co/c7K3O0QZ8o #TriAttention #LLM #KVCache #AIInference #VideoGeneration #vLLM #SGLang #LongLive #AIEfficiency #MLSystem #GPU #NVIDA

1

83

12

45

6K

0

2

0

138

Sony AI @SonyAI_global

4 months ago

EnTruth offers a new way to trace unauthorized dataset use in #diffusionmodels with minimal image alteration. 👉 Full paper: https://t.co/DrgYdglwSN

SonyAI_global's tweet photo. EnTruth offers a new way to trace unauthorized dataset use in #diffusionmodels with minimal image alteration. 👉 Full paper: https://t.co/DrgYdglwSN https://t.co/UJB0BsqBTQ

0

2

1

2

437

Chieh-Hsin (Jesse) Lai

@JCJesseLai

5 months ago

🌍 Website: https://t.co/w9y0ENTmw3 🗓️ Schedule: https://t.co/4sAb2Sk7kY See you at CVPR! #CVPR2026 #DiffusionModels #DiffusionLLMs

0

9

3

1K

Subham Sahoo

@ssahoo_

5 months ago

🌍 Website: https://t.co/s8W0bHhSKD 📅 Schedule: https://t.co/usY0ihpvYw See you at CVPR! #CVPR2026 #DiffusionModels #DiffusionLLMs (3/3) (3 / 3)

1

18

2

10

1K

TruthSeeker

@MoSphere420

5 months ago

Why process noise at full res? Our paper reveals diffusion's hidden info hierarchy: highly noisy states are just tiny images. We fuse scale spaces to fix this massive compute inefficiency. Rethink how AI generates. #AI #DiffusionModels https://t.co/mKefhKk7mV

0

3

0

26

Eduardo Alvarado @roboonaut

5 months ago

Our multi-modal diffusion architecture captures a massive range of movement—from walking and jogging to crouching, tiptoeing, and even dancing!🕺 All of this is reconstructed using only 16 pressure sensors and an IMU per insole. #MotionCapture #Wearables #DiffusionModels

1

4

0

471

Sony AI @SonyAI_global

5 months ago

🧠📖A thoughtful read for researchers and practitioners 🔗https://t.co/I5bt0tSmUV #DiffusionModels #GenerativeAI #AIResearch #MachineLearning

0

2

1

0

239

Evolution AI Hub @EvolutionAIHub

5 months ago

Autoregressive LLMs: types carefully one word at a time Mercury 2: throws the entire dictionary at the wall and refines it instantly 1,196 tokens/sec is absurd. 🤯 🚀 #Mercury2 #DiffusionModels #AIart https://t.co/2VB4WrOuMw

0

2

0

58

deeplearningforyou @deeplearningxd

5 months ago

🤯 Why did AI ditch GANs for diffusion models? GANs: Unstable training + mode collapse Diffusion: Stable denoising + diverse outputs That's why DALL-E, Stable Diffusion & Midjourney all chose diffusion! 🎥 Watch the full breakdown: #AI #DeepLearning #DiffusionModels #GenerativeAI

1

2

0

56

Top Tweets for #Diffusionmodels

Last Seen Hashtags on Sotwe

Trends for you

Most Popular Users