Predicting Mergeability of Parameter-Efficient Fine-Tuning Updates

2026-06-17 · Lin Tang, Wei Zhang, Jing Li, Hongyu Chen, Ming Zhao, Yuxuan Wang arxiv

Low-rank adaptation (LoRA) makes it cheap to train many domain- and task-specific language model adapters, but whether two adapters can be merged is usually discovered only after both have been fully trained and evaluated. This late feedback is costly: adapters that are strong in isolation can interfere destructively once their updates are combined. We ask whether this outcome can be anticipated. We formalize adapter mergeability as the degree to which an adapter preserves its single-task utility after merging, and show that it can be forecast from signals measured in the first few percent of training -- chiefly how the low-rank updates and their gradients align across tasks and how much they disturb shared representations. We package these signals into MergeProbe, a lightweight predictor that estimates pairwise and set-level retention and turns the estimate into a concrete decision: merge directly, reweight, prune, or route. On MERGE-PEFT, a five-domain benchmark spanning math, code, science, instruction following, and safety, MergeProbe attains the best average and worst-case retention among strong interference-aware merge baselines while adding far less deployment overhead than full task routing. This turns LoRA merging from a post-hoc engineering step into an anticipatory measurement problem.

📄 PDF Abstract BibTeX arXiv:2606.19549

Code (0)

등록된 구현이 없습니다.

Tasks

parameter-efficient fine-tuning Instruction Following

Demystifying Mergeability: Interpretable Properties to Predict Model Merging Success

2026-01-29 · Luca Zhou, Bo Zhao, Rose Yu, Emanuele Rodolà arxiv

Model merging combines knowledge from separately fine-tuned models, yet the factors driving its success remain poorly understood. While recent work treats mergeability as an intrinsic property of the models, we show with…

Will it Merge? On The Causes of Model Mergeability

2026-01-10 · Adir Rahamim, Asaf Yehudai, Boaz Carmeli, Leshem Choshen 외 arxiv

Model merging has emerged as a promising technique for combining multiple fine-tuned models into a single multitask model without retraining. However, the factors that determine whether merging will succeed or fail remai…

MergeVLA: Cross-Skill Model Merging Toward a Generalist Vision-Language-Action Agent

2025-11-24 · Yuxia Fu, Zhizhen Zhang, Yuqi Zhang, Zijian Wang 외 arxiv

Recent Vision-Language-Action (VLA) models reformulate vision-language models by tuning them with millions of robotic demonstrations. While they perform well when fine-tuned for a single embodiment or task family, extend…

AFA-LoRA: Enabling Non-Linear Adaptations in LoRA with Activation Function Annealing

2025-12-27 · Jiacheng Li, Jianchao Tan, Zhidong Yang, Feiye Huo 외 arxiv

Low-Rank Adaptation (LoRA) is a widely adopted parameter-efficient fine-tuning (PEFT) method. However, its linear adaptation process limits its expressive power. This means there is a gap between the expressive power of …

parameter-efficient fine-tuning Reinforcement Learning

When Privacy Hurts Mergeability: Geometry-Aware Model Merging under Differential Privacy

2026-08-27 · Jin Liu, Junkang Liu, Ning Xi, Yinbin Miao 외 arxiv

Model merging promises to construct a single multi-task model from independently fine-tuned task models without accessing the original task data. This makes it attractive when task data cannot be centralized, but release…