Cached Multi-Lora Composition for Multi-Concept Image Generation
Xiandong Zou, Mingzhu Shen, Christos-Savvas Bouganis, Yiren Zhao
Abstract
Low-Rank Adaptation (LoRA) has emerged as a widely adopted technique in textto-image models, enabling precise rendering of multiple distinct elements, such as characters and styles, in multi-concept image generation. However, current approaches face significant challenges when composing these LoRAs for multiconcept image generation, particularly as the number of LoRAs increases, resulting in diminished generated image quality. In this paper, we initially investigate the role of LoRAs in the denoising process through the lens of the Fourier frequency domain. Based on the hypothesis that applying multiple LoRAs could lead to "semantic conflicts", we have conducted empirical experiments and find that certain LoRAs amplify high-frequency features such as edges and textures, whereas others mainly focus on low-frequency elements, including the overall structure and smooth color gradients. Building on these insights, we devise a frequency domain based sequencing strategy to determine the optimal order in which LoRAs should be integrated during inference. This strategy offers a methodical and generalizable solution compared to the naive integration commonly found in existing LoRA fusion techniques. To fully leverage our proposed LoRA order sequence determination method in multi-LoRA composition tasks, we introduce a novel, training-free framework, Cached Multi-LoRA (CMLoRA), designed to efficiently integrate multiple LoRAs while maintaining cohesive image generation. With its flexible backbone for multi-LoRA fusion and a non-uniform caching strategy tailored to individual LoRAs, CMLoRA has the potential to reduce semantic conflicts in LoRA composition and improve computational efficiency. Our experimental evaluations demonstrate that CMLoRA outperforms state-of-the-art training-free LoRA fusion methods by a significant margin -it achieves an average improvement of 2.19% in CLIPScore, and 11.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
5 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- High-Resolution Image Synthesis with Latent Diffusion Models cites Zero-Shot Text-to-Image Generation
- High-Resolution Image Synthesis with Latent Diffusion Models cites Denoising Diffusion Probabilistic Models
- QR-LoRA: Efficient and Disentangled Fine-Tuning via QR Decomposition for Customized Generation cites LoRA: Low-Rank Adaptation of Large Language Models
- QR-LoRA: Efficient and Disentangled Fine-Tuning via QR Decomposition for Customized Generation cites High-Resolution Image Synthesis with Latent Diffusion Models
- QR-LoRA: Efficient and Disentangled Fine-Tuning via QR Decomposition for Customized Generation cites Denoising Diffusion Probabilistic Models
- Diffusion Blend: Inference-Time Multi-Preference Alignment for Diffusion Models cites High-Resolution Image Synthesis with Latent Diffusion Models
- Diffusion Blend: Inference-Time Multi-Preference Alignment for Diffusion Models cites Denoising Diffusion Probabilistic Models
- Preference-Aligned LoRA Merging: Preserving Subspace Coverage and Addressing Directional Anisotropy cites LoRA: Low-Rank Adaptation of Large Language Models
- SG-LoRA: Semantic-guided LoRA Parameters Generation cites LoRA: Low-Rank Adaptation of Large Language Models
- SG-LoRA: Semantic-guided LoRA Parameters Generation cites High-Resolution Image Synthesis with Latent Diffusion Models
- SG-LoRA: Semantic-guided LoRA Parameters Generation cites Denoising Diffusion Probabilistic Models
- Directly Denoising Diffusion Models cites High-Resolution Image Synthesis with Latent Diffusion Models
- Directly Denoising Diffusion Models cites Denoising Diffusion Probabilistic Models
Cited by top-tier papers5
- Diffusion Blend: Inference-Time Multi-Preference Alignment for Diffusion ModelsMin Cheng, Fatemeh Doudi, Dileep Kalathil, Mohammad Ghavamzadeh et al.ICLR 2026 · 6 citations
- QR-LoRA: Efficient and Disentangled Fine-Tuning via QR Decomposition for Customized GenerationJiahui Yang, Yongjia Ma, Donglin Di, Jianxun Cui et al.ICCV 2025 · 3 citations
- Preference-Aligned LoRA Merging: Preserving Subspace Coverage and Addressing Directional AnisotropyWooseong Jeong, Wonyoung Lee, Kuk-Jin YoonCVPR 2026 · 1 citation
- VMDiff: Visual Mixing Diffusion for Limitless Cross-Object SynthesisZeren Xiong, Yue Yu, Ze-dong Zhang, Shuo Chen et al.ICLR 2026 · 1 citation
- SG-LoRA: Semantic-guided LoRA Parameters GenerationMiaoge Li, Yang Chen, Zhijie Rao, Can Jiang et al.CVPR 2026
Builds on19
- Denoising Diffusion Probabilistic ModelsJonathan Ho, Ajay Jain, Pieter AbbeelNeurIPS 2020 · 35,902 citations
- LoRA: Low-Rank Adaptation of Large Language ModelsEdward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu et al.ICLR 2022 · 18,833 citations
- High-Resolution Image Synthesis with Latent Diffusion ModelsRobin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser et al.CVPR 2022 · 13,123 citations
- Directly Denoising Diffusion ModelsDan Zhang, Jingjing Wang, Feng LuoICML 2024 · 11,724 citations
- Zero-Shot Text-to-Image GenerationAditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray et al.ICML 2021 · 6,356 citations
Related papers
- Contrastive Test-Time Composition of Multiple LoRA Models for Image GenerationTuna Han Salih Meral, Enis Simsar, Federico Tombari, Pinar YanardagICCV 2025 · 1 citation
- TARA: Token-Aware LoRA for Composable Personalization in Diffusion ModelsYuqi Peng, Lingtao Zheng, Yufeng Yang, Yi Huang et al.AAAI 2026 · 2 citations
- CRAFT-LoRA: Content-Style Personalization via Rank-Constrained Adaptation and Training-Free FusionYu Li, Yujun Cai, Chi ZhangCVPR 2026 · 2 citations
- ELORA: Efficient LoRA and KV Cache Management for Multi-LoRA LLM ServingJiuchen Shi, Hang Zhang, Yixiao Wang, Quan Chen et al.HPCA 2026
- LoRA on the Go: Instance-level Dynamic LoRA Selection and MergingSeungeon Lee, Soumi Das, Manish Gupta, Krishna P. GummadiACL 2026 · 2 citations
