← 研踪 PaperTrailarXiv:2611.21077 · v1
MORE: Merge Once, Route Early for Training-Free Multi-Subject Composition from Large LoRA Pools
首次提交 2026-11-16 21:28
版本更新 2026-11-16 21:28
雷达观测 2026-11-17 17:11
摘要
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 or 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% () and 78.3% () 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 subjects stay in one adapter. With layouts inferred automatically it still keeps both subjects in 70.0% / 55.0% of images.
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-16 21:28 · 最新观测版本
MORE: Merge Once, Route Early for Training-Free Multi-Subject Composition from Large LoRA Pools
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 or 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% () and 78.3% () 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 subjects stay in one adapter. With layouts inferred automatically it still keeps both subjects in 70.0% / 55.0% of images.
查看此版本解读 →查看此版本原文 ↗