Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

RSTFA: Efficient Training-Free Human-Preference Alignment via Rejection Sampling for Text-to-Image Diffusion Models.

Hongzheng Yang, Jason Chun-Lok Li, Li Kun

    IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society
    |April 21, 2026
    PubMed
    Summary
    This summary is machine-generated.

    Related Concept Videos

    Improving Translational Accuracy02:07

    Improving Translational Accuracy

    2.6K
    2.6K
    Improving Translational Accuracy02:07

    Improving Translational Accuracy

    11.5K
    Base complementarity between the three base pairs of mRNA codon and the tRNA anticodon is not a failsafe mechanism. Inaccuracies can range from a single mismatch to no correct base pairing at all. The free energy difference between the correct and nearly correct base pairs can be as small as 3 kcal/ mol. With complementarity being the only proofreading step, the estimated error frequency would be one wrong amino acid in every 100 amino acids incorporated. However, error frequencies observed in...
    11.5K

    You might also read

    Related Articles

    Articles linked to this work by shared authors, journal, and citation graph.

    Sort by
    Same author

    An interpretable memory forensics framework for unknown attack identification in power grid edge devices.

    Scientific reports·2026
    Same author

    DRID: a spatiotemporal relational framework for robust IoT device identification in smart grids.

    Scientific reports·2026
    Same author

    Ophthalmomyiasis Outbreak Caused by Oestrus ovis Infection, Algeria, 2025.

    Emerging infectious diseases·2026
    Same author

    Primary multifocal gastrointestinal epithelioid angiosarcoma with MYC amplification: a diagnostic pitfall mimicking poorly differentiated carcinoma: a case report.

    BMC gastroenterology·2026
    Same author

    Dietary 3,4-Dihydroxybenzalacetone alleviates LPS-induced intestinal injury in Leizhou black ducks by improving mitochondrial function and suppressing apoptosis.

    Poultry science·2026
    Same author

    Neural Network Enabled Real-Time Plasma Imaging for Inverse Designed Fabrication of Micro/Nano Structures with Ultrafast Laser.

    Advanced materials (Deerfield Beach, Fla.)·2026
    Same journal

    Hyperbolic Cycle Alignment for Infrared-Visible Image Fusion.

    IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
    Same journal

    Learning Gaze Synthesizer via 3D-eye Controlled Diffusion and Cross-domain Feature Alignment.

    IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
    Same journal

    Underlying Semantic Diffusion for Effective and Efficient In-Context Learning.

    IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
    Same journal

    DiffRES: Unleashing Text-to-Image Diffusion Models for Generative Referring Expression Segmentation without Information Leakage.

    IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
    Same journal

    Location Matters: Frequency-Spatial Dual Space Adaptation for Cross-Domain Few-Shot Segmentation.

    IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
    Same journal

    BayeTopo: Bayesian-based Topology-guided Learning for Vascular Imaging Segmentation.

    IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
    See all related articles

    This study reveals that aligning text-to-image diffusion models mainly impacts style, not content. A new training-free method, Rejection Sampling Training-Free Alignment (RSTFA), efficiently aligns models with human preferences without fine-tuning.

    Area of Science:

    • Artificial Intelligence
    • Computer Vision
    • Machine Learning

    Background:

    • Text-to-image diffusion models are powerful generative tools.
    • Aligning these models with human preferences is crucial for usability.
    • Current alignment methods often require extensive fine-tuning.

    Purpose of the Study:

    • To investigate the effects of alignment tuning on diffusion models.
    • To develop a training-free alignment method that preserves model diversity and reduces computational cost.
    • To demonstrate superior alignment with human preferences compared to existing methods.

    Main Methods:

    • Analysis of denoising trajectories in base and aligned diffusion models.
    • Introduction of Rejection Sampling Training-Free Alignment (RSTFA) using rejection sampling at stylistic timesteps.

    Related Experiment Videos

  • Theoretical analysis and bias bound derivation for the RSTFA scheme.

    • Main Results:

    • Alignment tuning primarily affects superficial stylistic aspects, not core content.
    • RSTFA achieves human preference alignment without fine-tuning or significant inference overhead.
    • RSTFA preserves sample diversity better than reinforcement-learning-based tuning.
    • Experiments show RSTFA outperforms state-of-the-art methods on multiple benchmarks.

    Conclusions:

    • Diffusion model alignment can be achieved efficiently and without training.
    • RSTFA offers a computationally inexpensive and effective approach to human-centered AI.
    • The findings pave the way for more accessible and user-aligned generative models.