Diffusion Transformers
PulseAugur coverage of Diffusion Transformers — every cluster mentioning Diffusion Transformers across labs, papers, and developer communities, ranked by signal.
- instance of Diffusion Transformer 90%
- instance of CatalyzeX Code Finder for Papers 90%
- instance of Pixart 90%
- instance of Diffusion Models 90%
- instance of Qwen Image 90%
- developed Data Influence Oriented Tree Search 90%
- used by Wan2.1-14B 90%
- instance of DiT-XL/2 90%
- used by Data Influence Oriented Tree Search 70%
- used by HunyuanVideo 70%
- used by Diffusion Models 70%
- used by Diffusion Transformer 70%
9 day(s) with sentiment data
-
New DiRotQ method enhances 4-bit quantization for Diffusion Transformers
Researchers have developed DiRotQ, a new 4-bit quantization framework for Diffusion Transformers (DiTs) that significantly reduces memory and computational costs while preserving image generation quality. The method uti…
-
New method detects and suppresses risky content in Diffusion Transformers
Researchers have developed a novel, training-free method to detect and suppress risky content generated by Diffusion Transformers (DiTs), a type of text-to-image model. The technique identifies concepts within the model…
-
InstaGen accelerates video generation with new sampling technique
Researchers have developed InstaGen, a novel method to accelerate video generation using Diffusion Transformers (DiTs). InstaGen addresses the computational cost of DiTs by employing group-shared importance sampling, wh…
-
DualCast framework enhances digital human e-commerce video generation
Researchers have introduced DualCast, a novel framework designed to improve the generation of digital human e-commerce videos. This method explicitly addresses the distinct roles of product appearance and human identity…
-
RouTex framework advances 3D texture generation using reprogrammed positional encoding · 2 sources tracked
Researchers have introduced RouTex, a new framework for generating 3D textures using multi-view reference images. The system leverages in-context Diffusion Transformers and reprograms 3D Rotary Position Encoding (RoPE) …
-
New theories link consciousness to quantum mechanics and AI models
Two related posts on Mastodon explore novel theories about consciousness. One proposes that human consciousness mirrors quantum mechanics, with time unfolding in discrete moments and information loops. The other introdu…
-
SAKURA architecture enables private image editing on phones
Researchers have developed SAKURA, a cloud-edge architecture designed to efficiently serve attention key-value (KV) states for private image editing on mobile devices. This system addresses the challenge of storing larg…
-
New LSGQuant method enhances diffusion models for video super-resolution
Researchers have developed LSGQuant, a novel quantization method designed to reduce the computational cost and model size of one-step diffusion models used for real-world video super-resolution. This approach incorporat…
-
PISA method boosts Diffusion Transformer efficiency with piecewise sparse attention
Researchers have developed PISA, a new training-free method for Piecewise Sparse Attention designed to improve the efficiency of Diffusion Transformers. PISA addresses the quadratic complexity of attention in these mode…
-
New AI models tackle unrestricted image resolution and agricultural robot navigation
Researchers have introduced FiTv2, an improved Flexible Vision Transformer designed for diffusion models, capable of generating images at unrestricted resolutions and aspect ratios. This architecture addresses limitatio…
-
TamedFlow accelerates diffusion model sampling by bridging compression and adaptive inference
Researchers have developed TamedFlow, a novel method for accelerating diffusion model sampling by combining model compression with adaptive inference techniques. This approach addresses the "Combination Gap," a quality …
-
MoDDiT enhances Diffusion Transformer efficiency with dynamic token routing
Researchers have introduced MoDDiT, a novel architecture modification for Diffusion Transformers (DiTs) designed to enhance efficiency in image and video generation. MoDDiT enables dynamic computation allocation by lear…
-
SpecKCast accelerates diffusion transformers using frequency-domain Kalman filtering
Researchers have developed SpecKCast, a novel training-free method designed to accelerate diffusion transformers, which are commonly used for image and video generation. This technique forecasts the transformer's output…
-
UniW(A)M framework unifies world and action models using diffusion Transformers
Researchers have introduced UniW(A)M, a novel mixture-of-experts framework designed to unify world models and world-action models. This framework utilizes diffusion Transformers to share common physical dynamics while a…
-
New MOSAIC framework enhances diffusion Transformers for image super-resolution
Researchers have developed MOSAIC, a novel framework designed to adapt pretrained diffusion Transformers for structure-sensitive image restoration tasks. This object-centric approach organizes degraded images into struc…
-
Diffusion Transformers' Representation Learning Weakened Without Long Skip Connections
A new research paper published on arXiv explores the performance of Diffusion Transformers (DiT) as representation learners. The study finds that DiT models, despite their strong generative capabilities, tend to produce…
-
New DAGS method enhances DiT image generation with temporal stability
Researchers have developed DAGS, a novel conditioning scheme for Diffusion Transformers (DiTs) that enhances image generation quality and temporal stability. This method uses lightweight, attention-free encoders to stee…
-
New framework GeoSim analyzes VLM representations for image restoration
Researchers have developed a new framework called GeoSim to analyze the internal representations of vision-language models (VLMs) for low-level image restoration tasks. The study investigates how different VLM architect…
-
Diffusion models advance with new architectures and training methods · 10 sources tracked
Researchers are exploring advanced diffusion models across various domains, including tri-modal generation for text, images, and audio by Apple, and data-free training methods for lattice field theories using diffusion …
-
New research tackles video diffusion model efficiency and safety
Researchers are developing new methods to improve the efficiency and quality of video diffusion models. SplitMoE introduces a split-role architecture to better handle the semantic imbalance in video data, outperforming …