R

Reasongen R1 SFT

Developed by Franklin0
ReasonGen-R1 is a text-to-image model trained on image prompts and reasoning basis datasets through supervised fine-tuning (SFT), with the explicit 'thinking' ability based on text.
Downloads 312
Release Time : 5/27/2025

Model Overview

ReasonGen-R1 is a two-stage framework. First, the autoregressive image generator is equipped with text-based reasoning ability through supervised fine-tuning, and then the output is optimized using Group Relative Policy Optimization (GRPO).

Model Features

Chain-of-thought reasoning ability
Through supervised fine-tuning, the model is equipped with the explicit 'thinking' ability based on text and can perform controllable planning of object layout, style, and scene combination.
Two-stage optimization framework
First, perform supervised fine-tuning (SFT), and then use Group Relative Policy Optimization (GRPO) to optimize the output.
Automatically generated reasoning basis corpus
Publish a model-generated reasoning basis corpus paired with visual prompts to support controllable image generation planning.

Model Capabilities

Text-to-image generation
Text-based reasoning
Controllable image planning

Use Cases

Creative design
Scene design
Generate complex scene layouts according to text descriptions
Generate detailed scene images that meet the text reasoning basis
Stylized image generation
Generate images in a specific artistic style based on style descriptions
Generate artworks with consistent style and meeting expectations
Education
Visual teaching material generation
Generate supporting visual materials according to teaching needs
Generate images highly relevant to the teaching content
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase