CoCoA-Mix: Confusion-and-Confidence-Aware Mixture Model for Context Optimization

Dasol Hong, Wooju Lee, Hyun Myung
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:23700-23721, 2025.

Abstract

Prompt tuning, which adapts vision-language models by freezing model parameters and opti- mizing only the prompt, has proven effective for task-specific adaptations. The core challenge in prompt tuning is improving specialization for a specific task and generalization for unseen domains. However, frozen encoders often produce misaligned features, leading to confusion between classes and limiting specialization. To overcome this issue, we propose a confusion-aware loss (CoA-loss) that improves specialization by refining the decision boundaries between confusing classes. Additionally, we mathematically demonstrate that a mixture model can enhance generalization without compromising specialization. This is achieved using confidence-aware weights (CoA- weights), which adjust the weights of each prediction in the mixture model based on its confidence within the class domains. Extensive experiments show that CoCoA-Mix, a mixture model with CoA-loss and CoA-weights, outperforms state-of-the-art methods by enhancing specialization and generalization. Our code is publicly available at https://github.com/url-kaist/CoCoA-Mix

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-hong25e, title = {{C}o{C}o{A}-Mix: Confusion-and-Confidence-Aware Mixture Model for Context Optimization}, author = {Hong, Dasol and Lee, Wooju and Myung, Hyun}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {23700--23721}, 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/hong25e/hong25e.pdf}, url = {https://proceedings.mlr.press/v267/hong25e.html}, abstract = {Prompt tuning, which adapts vision-language models by freezing model parameters and opti- mizing only the prompt, has proven effective for task-specific adaptations. The core challenge in prompt tuning is improving specialization for a specific task and generalization for unseen domains. However, frozen encoders often produce misaligned features, leading to confusion between classes and limiting specialization. To overcome this issue, we propose a confusion-aware loss (CoA-loss) that improves specialization by refining the decision boundaries between confusing classes. Additionally, we mathematically demonstrate that a mixture model can enhance generalization without compromising specialization. This is achieved using confidence-aware weights (CoA- weights), which adjust the weights of each prediction in the mixture model based on its confidence within the class domains. Extensive experiments show that CoCoA-Mix, a mixture model with CoA-loss and CoA-weights, outperforms state-of-the-art methods by enhancing specialization and generalization. Our code is publicly available at https://github.com/url-kaist/CoCoA-Mix} }
Endnote
%0 Conference Paper %T CoCoA-Mix: Confusion-and-Confidence-Aware Mixture Model for Context Optimization %A Dasol Hong %A Wooju Lee %A Hyun Myung %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-hong25e %I PMLR %P 23700--23721 %U https://proceedings.mlr.press/v267/hong25e.html %V 267 %X Prompt tuning, which adapts vision-language models by freezing model parameters and opti- mizing only the prompt, has proven effective for task-specific adaptations. The core challenge in prompt tuning is improving specialization for a specific task and generalization for unseen domains. However, frozen encoders often produce misaligned features, leading to confusion between classes and limiting specialization. To overcome this issue, we propose a confusion-aware loss (CoA-loss) that improves specialization by refining the decision boundaries between confusing classes. Additionally, we mathematically demonstrate that a mixture model can enhance generalization without compromising specialization. This is achieved using confidence-aware weights (CoA- weights), which adjust the weights of each prediction in the mixture model based on its confidence within the class domains. Extensive experiments show that CoCoA-Mix, a mixture model with CoA-loss and CoA-weights, outperforms state-of-the-art methods by enhancing specialization and generalization. Our code is publicly available at https://github.com/url-kaist/CoCoA-Mix
APA
Hong, D., Lee, W. & Myung, H.. (2025). CoCoA-Mix: Confusion-and-Confidence-Aware Mixture Model for Context Optimization. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:23700-23721 Available from https://proceedings.mlr.press/v267/hong25e.html.

Related Material