Wiener-informed decoding speeds pixel-space flow model training
On ImageNet 256, PixelDiT-XL with WiD reports FID 1.53 after 500 epochs, matching prior quality with 37.5% fewer epochs.
Stories per day · 30 days
Relatedautoregressive1diffusion1image-generation1imagenet1lafa1language-models1pixel-space1speculative-decoding1
7 stories, newest first · one card per story
Stories tagged #flow-models
Lafa scales continuous flow language models to 8B
Trained on fewer than 30B tokens, its self-speculative decoding reaches 2.54 times autoregressive throughput at comparable quality.
InstantFusion: A Shared Latent Interface for Heterogeneous Image Generators
InstantFusion connects heterogeneous image generators through a shared latent denoising interface.
Data-free method distills diffusion models to one-step generation
Authors report ImageNet 256 FID of 2.04 at 1-NFE and 1.37 at 4-NFE after 40 epochs, without distillation data.
MEND: RL For Flow Models via Proximal Velocity Matching
MEND post-trains flow models with proximal velocity matching, beating Flow-GRPO and ReFL in far fewer updates.
Latent-Foresight: End-to-End Learning Predictable Representations for Latent World Models
Latent-Foresight jointly learns a tokenizer and flow model so vision latents stay predictable.
Discrete Beckmann Transport Models for One-Step Language Modeling and Reasoning
DBTM achieves one-step text generation via a time-independent transport map trained directly from data, removing pretrained teacher distillation.