Images Talk Back: Guiding MMDiTs by Suppressing Image-to-Text Feedback

cs.CV

Submitted: 2026-10-09

Updated: 2026-10-12

Comments: 26 pages, including appendix

Code: https://github.com/yyc1215lll-creator/SFG

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• CFG-Zero*: Improved Classifier-Free Guidance for Flow Matching Models
• Vision-Language Binding in In-Context Image Generation
• Classifier-Free Diffusion Guidance
• HunyuanVideo: A Systematic Framework For Large Video Generative Models
• Text Template Tokens Are Implicit Semantic Registers in Diffusion Transformers
• Playground v3: Improving Text-to-Image Alignment with Deep-Fusion Large Language Models
• DifFRACT: Diffusion Feature Reconstruction and Attribution for Circuit Tracing
• Qwen-Image Technical Report
• Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis
• 3DIS-FLUX: simple and efficient multi-instance generation with DiT rendering

Related papers

• Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
• AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
• Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
• MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
• TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
• A Survey on Efficient Vision-Language-Action Models