Papers
Sign in to view your remaining parses.
Tag Filter
High-Fidelity Image Synthesis
NextStep-1: Toward Autoregressive Image Generation with Continuous Tokens at Scale
Published:8/14/2025
Autoregressive Image GenerationText-to-Image GenerationHigh-Fidelity Image SynthesisContinuous Image Token ProcessingDiscrete Text Token Training
NextStep1 is an autoregressive model that effectively handles continuous image tokens without relying on heavy diffusion models or incurring quantization loss. It achieves stateoftheart performance in texttoimage generation and excels in image editing, demonstrating its pow
02
One Small Step in Latent, One Giant Leap for Pixels: Fast Latent Upscale Adapter for Your Diffusion Models
Published:11/14/2025
Diffusion Model Super-ResolutionLightweight Latent Upscaler AdapterSwin-Style BackboneHigh-Fidelity Image SynthesisLatent Space Image Generation
The paper introduces a lightweight Latent Upscaler Adapter (LUA) that performs superresolution directly on latent codes in diffusion models, reducing image generation time by nearly threefold while maintaining comparable perceptual quality, facilitating highfidelity synthesis w
02