CT-Merging: Consensus Directions and Task-Specific Scaling for LoRA Adapter Merging

arXiv:2607.20561 · cs.LG, cs.CV · Submitted 2026-07-18 · Read on arXiv

cs.LG, cs.CV

Submitted: 2026-07-18

Updated: 2026-09-25

Comments: 5 pages, 1 figure

License: http://creativecommons.org/licenses/by-sa/4.0/

The gist: LoRA merging methods increasingly operate on the low-rank structure of task updates, yet how the common subspace is estimated and how coefficients are assigned after recomposition are rarely compared

Terminology

Abstract

LoRA merging methods increasingly operate on the low-rank structure of task updates, yet how the common subspace is estimated and how coefficients are assigned after recomposition are rarely compared directly. We propose CT-Merging, which estimates common directions from averaged task subspace projectors and assigns a separate residual scale to each task. Projector averaging selects directions supported across task subspaces without weighting them by singular magnitude, while task-specific scaling removes component-wise magnitude variation and preserves scale differences across tasks. On the released KnOTS CLIP adapters, CT-Merging achieves the best average and worst-task normalized accuracy on both backbones, improving over the strongest baseline by up to 2.56 and 6.65 points, respectively. On the DC-Merge adapter benchmark, it achieves the best average normalized accuracy in eight of nine backbone and task-count settings. Ablations show that projector averaging outperforms summed-update SVD and that task-specific scaling improves worst-task accuracy over global isotropic scaling.

Sources

Related papers