CAT Merging: A Training-Free Approach for Resolving Conflicts in Model Merging

Wenju Sun, Qingyong Li, Yangliao Geng, Boyang Li
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:57523-57543, 2025.

Abstract

Multi-task model merging offers a promising paradigm for integrating multiple expert models into a unified system without additional training. Existing state-of-the-art techniques, such as Task Arithmetic and its variants, merge models by accumulating task vectors—defined as the parameter differences between pre-trained and fine-tuned models. However, task vector accumulation is often hindered by knowledge conflicts, where conflicting components across different task vectors can lead to performance degradation during the merging process. To address this challenge, we propose Conflict-Aware Task Merging (CAT Merging), a novel training-free framework that selectively trims conflict-prone components from the task vectors. CAT Merging introduces several parameter-specific strategies, including projection for linear weights and masking for scaling and shifting parameters in normalization layers. Extensive experiments on vision and vision-language tasks demonstrate that CAT Merging effectively suppresses knowledge conflicts, achieving average accuracy improvements of up to 4.7% (ViT-B/32) and 2.0% (ViT-L/14) over state-of-the-art methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-sun25i, title = {{CAT} Merging: A Training-Free Approach for Resolving Conflicts in Model Merging}, author = {Sun, Wenju and Li, Qingyong and Geng, Yangliao and Li, Boyang}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {57523--57543}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/sun25i/sun25i.pdf}, url = {https://proceedings.mlr.press/v267/sun25i.html}, abstract = {Multi-task model merging offers a promising paradigm for integrating multiple expert models into a unified system without additional training. Existing state-of-the-art techniques, such as Task Arithmetic and its variants, merge models by accumulating task vectors—defined as the parameter differences between pre-trained and fine-tuned models. However, task vector accumulation is often hindered by knowledge conflicts, where conflicting components across different task vectors can lead to performance degradation during the merging process. To address this challenge, we propose Conflict-Aware Task Merging (CAT Merging), a novel training-free framework that selectively trims conflict-prone components from the task vectors. CAT Merging introduces several parameter-specific strategies, including projection for linear weights and masking for scaling and shifting parameters in normalization layers. Extensive experiments on vision and vision-language tasks demonstrate that CAT Merging effectively suppresses knowledge conflicts, achieving average accuracy improvements of up to 4.7% (ViT-B/32) and 2.0% (ViT-L/14) over state-of-the-art methods.} }
Endnote
%0 Conference Paper %T CAT Merging: A Training-Free Approach for Resolving Conflicts in Model Merging %A Wenju Sun %A Qingyong Li %A Yangliao Geng %A Boyang Li %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-sun25i %I PMLR %P 57523--57543 %U https://proceedings.mlr.press/v267/sun25i.html %V 267 %X Multi-task model merging offers a promising paradigm for integrating multiple expert models into a unified system without additional training. Existing state-of-the-art techniques, such as Task Arithmetic and its variants, merge models by accumulating task vectors—defined as the parameter differences between pre-trained and fine-tuned models. However, task vector accumulation is often hindered by knowledge conflicts, where conflicting components across different task vectors can lead to performance degradation during the merging process. To address this challenge, we propose Conflict-Aware Task Merging (CAT Merging), a novel training-free framework that selectively trims conflict-prone components from the task vectors. CAT Merging introduces several parameter-specific strategies, including projection for linear weights and masking for scaling and shifting parameters in normalization layers. Extensive experiments on vision and vision-language tasks demonstrate that CAT Merging effectively suppresses knowledge conflicts, achieving average accuracy improvements of up to 4.7% (ViT-B/32) and 2.0% (ViT-L/14) over state-of-the-art methods.
APA
Sun, W., Li, Q., Geng, Y. & Li, B.. (2025). CAT Merging: A Training-Free Approach for Resolving Conflicts in Model Merging. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:57523-57543 Available from https://proceedings.mlr.press/v267/sun25i.html.

Related Material