cs.CVNov 16, 2026

MORE: Merge Once, Route Early for Training-Free Multi-Subject Composition from Large LoRA Pools

Authors: Tian Xia

Abstract

Public hubs offer many personalized LoRAs per text-to-image backbone, yet composing two of them in one image remains fragile. In FLUX, keeping a pool of K=10K=10 or 2020 subject LoRAs loaded makes their deltas interfere, and even loading only the two requested LoRAs keeps both subjects in fewer than 4% of images: one subject vanishes, the two fuse, or the image splits into a collage. We present MORE (Merge Once, Route Early), a training-free two-stage pipeline for FLUX. Merge once: the whole pool is folded into a single fixed-rank adapter by the subspace-regression router of SSR-Merge, made reliable by ridge, rank and calibration choices, and factored exactly into a shared part and per-subject parts. Route early: during the first 30% of the denoising steps each subject's part is gated to its layout region; a decaying soft attention bias ties every region to its own subject tokens, and a cyclic swap guidance extrapolates away from the prediction with the subjects' regions permuted, which counteracts region homogenization. On held-out seeds, MORE keeps both subjects in 94.2% (K=10K=10) and 78.3% (K=20K=20) of images, against 50.8% / 35.8% for FreeFuse and 25.0% / 24.2% for LoRAShop, and reaches an ImageReward of 1.32 (best baseline 0.07), while all KK subjects stay in one adapter. With layouts inferred automatically it still keeps both subjects in 70.0% / 55.0% of images.

Figures & tables

Appendix figures & tables13 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

Oct 27, 2025

FreeFuse: Multi-Subject LoRA Fusion via Adaptive Token-Level Routing at Test Time

This paper proposes FreeFuse, a training-free framework for multi-subject text-to-image generation through automatic fusion of multiple subject LoRAs. In contrast to prior studies that focus on retraining LoRAs to alleviate feature conflicts, our analysis shows that spatially routing LoRA residuals to their intended semantic regions provides an effective mechanism for suppressing direct cross-region LoRA interference while preserving the base model's global contextual reasoning. Accordingly, we implement Adaptive Token-Level Routing during the inference phase. However, obtaining reliable routing regions remains challenging. Existing methods that rely on text-image latent association, such as raw cross-attention or concept-level similarity matching, often suffer from sparse activations, hole artifacts, and unstable localization when handling visually similar subjects, leading to incomplete or ambiguous subject masks. To address these issues, we introduce FreeFuseAttn, a mechanism that exploits the flow matching model's intrinsic semantic alignment to dynamically match subject-specific tokens to their corresponding spatial regions at early denoising timesteps, thereby bypassing the need for external segmentors. FreeFuse distinguishes itself through high practicality: it necessitates no additional training, model modifications, or user-defined spatial constraints. Users need only provide subject activation words to achieve seamless integration into standard workflows. Extensive experiments validate that FreeFuse outperforms existing approaches in both identity preservation and compositional fidelity. The project page is at: https://future-item.github.io/FreeFuse/.
Jun 2, 2026

Training-Free Multi-Concept LoRA Composition with Prompt-Aware Weighting

Low-Rank Adaptation (LoRA) successfully enables personalization in text-to-image generation by adapting pre-trained diffusion models to specific visual concepts and styles. However, extending such models to multi-concept customization remains challenging. Naively combining multiple LoRA weights or their outputs often leads to interference among concepts, resulting in degraded visual quality and reduced fidelity to the reference images of individual concepts. This paper proposes a simple yet effective approach for multi-concept customization by optimally combining the outputs of multiple LoRA modules. We leverage the relative importance of each concept during generation, as inferred from its corresponding prompt tokens and introduce two methods, W-Switch and W-Composite, that employ a prompt-aware importance weighting strategy in which each LoRA is weighted according to the semantic influence of its trigger words in the target prompt. In addition, we extend existing quantitative evaluation metrics by proposing a new image-based similarity evaluation framework that assesses image fidelity and identity preservation through comparisons between real-world reference images and automatically segmented concept regions from generated images. We evaluate our approach on the ComposLoRA testbed and demonstrate consistent improvements over existing state-of-the-art methods in terms of visual quality, identity preservation and compositionality. Qualitative evaluations, including a Large Language Model (LLM) based assessment and a user study, further validate the effectiveness of the proposed methods and align with the newly introduced quantitative image-based metrics. Our code is available at https://github.com/GeorgeTsoumplekas/Prompt-Aware-Multi-LoRA-Composition.
Sep 26, 2026

RefCompose: Multi-Reference Image Generation via LoRA-Conditioned Diffusion

Filmmakers and visual artists routinely need to compose multiple references, actors, locations, props, cultural elements, into a single coherent shot, but existing tools either fail to scale past a handful of references or destroy fine grained subject identity in the process, since per reference tokenization scales memory linearly with reference count NN and generated content often departs from the given references rather than reproducing them. We propose RefCompose, a pixel space compositional conditioning framework that decouples where things go from what they look like, via a single fixed resolution reference canvas that keeps conditioning size constant regardless of reference count. Spatial layout is induced at inference time from a frozen diffusion transformer and extracted via Grounding DINO, requiring no LLM or dedicated layout model, while dual stream LoRA adapters inject a layout derived depth map and the encoded canvas through separate low rank streams, disentangling geometric scaffolding from localized appearance. On the Dense Layout protocol, RefCompose consistently outperforms layout based and state of the art multi reference baselines on color, texture, shape, spatial accuracy, and identity/content preservation at higher reference counts, all with constant inference memory, making it a practical building block for multi subject cinematic composition at production scale.