Multi-Modality Expansion and Retention for LLMs through Parameter Merging and DecouplingAnnual Meeting of the Association for Computational Linguistics (ACL), 2025 |
LED-Merging: Mitigating Safety-Utility Conflicts in Model Merging with Location-Election-DisjointAnnual Meeting of the Association for Computational Linguistics (ACL), 2025 |
When, Where and Why to Average Weights?International Conference on Machine Learning (ICML), 2025 |
LoRA Soups: Merging LoRAs for Practical Skill Composition TasksInternational Conference on Computational Linguistics (COLING), 2024 |
Parameter Competition Balancing for Model MergingNeural Information Processing Systems (NeurIPS), 2024 |
DAM: Dynamic Adapter Merging for Continual Video QA LearningIEEE Workshop/Winter Conference on Applications of Computer Vision (WACV), 2024 |
A Quadratic Synchronization Rule for Distributed Deep LearningInternational Conference on Learning Representations (ICLR), 2023 |
AdaMerging: Adaptive Model Merging for Multi-Task LearningInternational Conference on Learning Representations (ICLR), 2023 |
The Split Matters: Flat Minima Methods for Improving the Performance of
GNNsInternational Cross-Domain Conference on Machine Learning and Knowledge Extraction (CD-MAKE), 2023 |
TIES-Merging: Resolving Interference When Merging ModelsNeural Information Processing Systems (NeurIPS), 2023 |
Understanding and Improving Model Averaging in Federated Learning on
Heterogeneous DataIEEE Transactions on Mobile Computing (IEEE TMC), 2023 |
Hierarchical Weight Averaging for Deep Neural NetworksIEEE Transactions on Neural Networks and Learning Systems (TNNLS), 2023 |
Randomized Adversarial Training via Taylor ExpansionComputer Vision and Pattern Recognition (CVPR), 2023 |
Dataless Knowledge Fusion by Merging Weights of Language ModelsInternational Conference on Learning Representations (ICLR), 2022 |
Diverse Weight Averaging for Out-of-Distribution GeneralizationNeural Information Processing Systems (NeurIPS), 2022 |
Consensus Control for Decentralized Deep LearningInternational Conference on Machine Learning (ICML), 2021 |