qa-FLoRA: Data-free query-adaptive Fusion of LoRAs for LLMs

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Shukla, Shreya, Sriram, Aditya, Narayanaswamy, Milinda Kuppur, Jain, Hiteshi
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911315724861440
author Shukla, Shreya
Sriram, Aditya
Narayanaswamy, Milinda Kuppur
Jain, Hiteshi
author_facet Shukla, Shreya
Sriram, Aditya
Narayanaswamy, Milinda Kuppur
Jain, Hiteshi
contents The deployment of large language models for specialized tasks often requires domain-specific parameter-efficient finetuning through Low-Rank Adaptation (LoRA) modules. However, effectively fusing these adapters to handle complex, multi-domain composite queries remains a critical challenge. Existing LoRA fusion approaches either use static weights, which assign equal relevance to each participating LoRA, or require data-intensive supervised training for every possible LoRA combination to obtain respective optimal fusion weights. We propose qa-FLoRA, a novel query-adaptive data-and-training-free method for LoRA fusion that dynamically computes layer-level fusion weights by measuring distributional divergence between the base model and respective adapters. Our approach eliminates the need for composite training data or domain-representative samples, making it readily applicable to existing adapter collections. Extensive experiments across nine multilingual composite tasks spanning mathematics, coding, and medical domains, show that qa-FLoRA outperforms static fusion by ~5% with LLaMA-2 and ~6% with LLaMA-3, and the training-free baselines by ~7% with LLaMA-2 and ~10% with LLaMA-3, while significantly closing the gap with supervised baselines. Further, layer-level analysis of our fusion weights reveals interpretable fusion patterns, demonstrating the effectiveness of our approach for robust multi-domain adaptation.
format Preprint
id arxiv_https___arxiv_org_abs_2512_11366
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle qa-FLoRA: Data-free query-adaptive Fusion of LoRAs for LLMs
Shukla, Shreya
Sriram, Aditya
Narayanaswamy, Milinda Kuppur
Jain, Hiteshi
Computation and Language
The deployment of large language models for specialized tasks often requires domain-specific parameter-efficient finetuning through Low-Rank Adaptation (LoRA) modules. However, effectively fusing these adapters to handle complex, multi-domain composite queries remains a critical challenge. Existing LoRA fusion approaches either use static weights, which assign equal relevance to each participating LoRA, or require data-intensive supervised training for every possible LoRA combination to obtain respective optimal fusion weights. We propose qa-FLoRA, a novel query-adaptive data-and-training-free method for LoRA fusion that dynamically computes layer-level fusion weights by measuring distributional divergence between the base model and respective adapters. Our approach eliminates the need for composite training data or domain-representative samples, making it readily applicable to existing adapter collections. Extensive experiments across nine multilingual composite tasks spanning mathematics, coding, and medical domains, show that qa-FLoRA outperforms static fusion by ~5% with LLaMA-2 and ~6% with LLaMA-3, and the training-free baselines by ~7% with LLaMA-2 and ~10% with LLaMA-3, while significantly closing the gap with supervised baselines. Further, layer-level analysis of our fusion weights reveals interpretable fusion patterns, demonstrating the effectiveness of our approach for robust multi-domain adaptation.
title qa-FLoRA: Data-free query-adaptive Fusion of LoRAs for LLMs
topic Computation and Language
url https://arxiv.org/abs/2512.11366