Flow Models
PulseAugur coverage of Flow Models — every cluster mentioning Flow Models across labs, papers, and developer communities, ranked by signal.
2 day(s) with sentiment data
-
New VAD and VA-TDM methods boost few-step diffusion model performance
Researchers have introduced Velocity Adversarial Distillation (VAD) and a new method called VA-TDM to improve the efficiency and quality of few-step diffusion and flow models. VAD uses a critic to compare student and te…
-
GazeCLIP model enhances deepfake detection using gaze vectors
Researchers have developed GazeCLIP, a novel system designed to improve the detection and attribution of deepfakes. This model leverages gaze vectors, which are observed to differ significantly between real and forged i…
-
InstantFusion connects heterogeneous image generators via shared latent interface
Researchers have developed InstantFusion, a novel shared latent interface designed to connect and translate between different image generation models. This interface leverages Latent AutoEncoders to project intermediate…
-
GS-Voxel framework enables scalable 3D scene generation from unstructured data
Researchers have developed GS-Voxel, a novel framework designed to convert unstructured 3D Gaussian Splatting (3DGS) reconstructions into structured latents. This method allows for scalable generation of large-scale 3D …
-
New algorithm aligns diffusion and flow models via black-box noise optimization
Researchers have developed a new trust-region based search algorithm (TRS) for aligning diffusion and flow models to specific rewards at inference time. This method treats the generative and reward models as black boxes…
-
Momentum Guidance enhances flow-based image generation quality
Researchers have introduced Momentum Guidance (MG), a new technique designed to enhance the quality of images generated by flow-based models. MG works by extrapolating the current velocity along the ODE trajectory, impr…
-
New self-guidance method boosts diversity in AI image generation
Researchers have developed a new training-free method called feature self-guidance to address diversity collapse in pretrained flow models used for image generation. This technique disperses internal features during bat…
-
LISA method accelerates AI model training for visual generation · 3 sources tracked
Researchers have introduced LISA (Likelihood Score Alignment), a novel regularization method designed to enhance the efficiency and performance of visual-condition controllable generation models. LISA works by explicitl…
-
FlowBender framework trains AI models to self-correct errors
Researchers have introduced FlowBender, a novel framework designed to improve the accuracy of conditional diffusion and flow models. This new approach trains models to utilize their own alignment errors as input, learni…
-
AI agents accelerate molecular design for batteries and drug discovery
Researchers have developed new AI agents for molecular design, aiming to accelerate scientific discovery. One agent, CLIO, uses a continuously updated belief-state graph and a recursive plan-then-act loop to recognize w…
-
NaviEdit improves image editing by decoupling model scale from edit progress
Researchers have developed NaviEdit, a new method to improve image editing with generative models. NaviEdit decouples the editing process from the model's scale, allowing for more semantic edits without sacrificing stru…