Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..

AlignedGen: Aligning Style Across Generated Images

Authors: Jiexuan Zhang, Yiheng Du, Qian Wang, Weiqi Li, Yu Gu, Jian Jun Zhang

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experimental results validate that our method effectively enhances style consistency across generated images while maintaining favorable text controllability. Code: https://github.com/Jiexuanz/Aligned Gen. 4 Experiment
Researcher Affiliation Academia 1School of Electronic and Computer Engineering, Peking University 2Guangdong Provincial Key Laboratory of Ultra High Definition Immersive Media Technology, Shenzhen Graduate School, Peking University EMAIL
Pseudocode Yes A Feature Extraction Pipeline Algorithm 1 Algorithm for Caching Q, K, V from a User-Provided Reference Image
Open Source Code Yes Code: https://github.com/Jiexuanz/Aligned Gen.
Open Datasets Yes Our evaluation dataset consists of 100 prompt sets from Style Aligned[16].
Dataset Splits Yes Our evaluation dataset consists of 100 prompt sets from Style Aligned[16].
Hardware Specification No The paper does not explicitly state specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments in Section 4.1 or any other part of the main text.
Software Dependencies Yes We apply our method on the Flux.1-dev [1]
Experiment Setup Yes For inference, we use the vanilla Rectified Flow sampler with 30 sampling steps and set the classifier-free guidance scale to 3.5. Specifically, setting ϕ to [19, 57) (the Single Blocks of Flux) yields optimal results.