A₃B₂: Adaptive Asymmetric Adapter for Alleviating Branch Bias in Vision-Language Image Classification with Few-Shot Learning

Zhou, Yiyun; Jiang, Zhonghua; Han, Wenkang; Li, Kunxi; Xu, Mingjing; Yao, Chang; Chen, Jingyuan

doi:10.24963/ijcai.2026/238

A₃B₂: Adaptive Asymmetric Adapter for Alleviating Branch Bias in Vision-Language Image Classification with Few-Shot Learning

Yiyun Zhou, Zhonghua Jiang, Wenkang Han, Kunxi Li, Mingjing Xu, Chang Yao, Jingyuan Chen

Download BibTeX Cite

Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence. Main Track. Pages 2136-2144.

https://doi.org/10.24963/ijcai.2026/238

Keywords

Computer Vision: Transfer, low-shot, semi- and un- supervised learning
Computer Vision: Multimodal learning

Abstract

Efficient transfer learning methods for large-scale vision–language models (e.g., CLIP) enable strong few-shot transfer, yet existing adaptation methods follow a fixed fine-tuning paradigm that implicitly assumes a uniform importance of the image and text branches, which has not been systematically studied in image classification. Through extensive analysis, we reveal a Branch Bias issue in vision–language image classification: adapting the image encoder does not always improve performance under out-of-distribution settings. Motivated by this observation, we propose A₃B₂, an Adaptive Asymmetric Adapter that alleviates Branch Bias in few-shot learning. A₃B₂ introduces Uncertainty-Aware Adapter Dampening (UAAD), which automatically suppresses image-branch adaptation when prediction uncertainty is high, enabling soft and data-driven control without manual intervention. Architecturally, A₃B₂ adopts a lightweight asymmetric design inspired by mixture-of-experts with Load Balancing Regularization. Extensive experiments demonstrate that A₃B₂ consistently outperforms 11 competitive prompt- and adapter-based baselines.

Yiyun Zhou, Zhonghua Jiang, Wenkang Han, Kunxi Li, Mingjing Xu, Chang Yao, and Jingyuan Chen. 2026. A₃B₂: Adaptive Asymmetric Adapter for Alleviating Branch Bias in Vision-Language Image Classification with Few-Shot Learning. In Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26. IJCAI, 2136–2144. https://doi.org/10.24963/ijcai.2026/238

@inproceedings{ijcai2026p238,
  title     = {A₃B₂: Adaptive Asymmetric Adapter for Alleviating Branch Bias in Vision-Language Image Classification with Few-Shot Learning},
  author    = {Zhou, Yiyun and Jiang, Zhonghua and Han, Wenkang and Li, Kunxi and Xu, Mingjing and Yao, Chang and Chen, Jingyuan},
  booktitle = {Proceedings of the Thirty-Fifth International Joint Conference on
               Artificial Intelligence, {IJCAI-26}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  editor    = {Diego Calvanese},
  pages     = {2136--2144},
  year      = {2026},
  month     = {8},
  note      = {Main Track},
  doi       = {10.24963/ijcai.2026/238},
  url       = {https://doi.org/10.24963/ijcai.2026/238},
}

TY  - CPAPER
TI  - A₃B₂: Adaptive Asymmetric Adapter for Alleviating Branch Bias in Vision-Language Image Classification with Few-Shot Learning
AU  - Zhou, Yiyun
AU  - Jiang, Zhonghua
AU  - Han, Wenkang
AU  - Li, Kunxi
AU  - Xu, Mingjing
AU  - Yao, Chang
AU  - Chen, Jingyuan
T2  - Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26
PB  - International Joint Conferences on Artificial Intelligence Organization
PY  - 2026
SP  - 2136
EP  - 2144
DO  - 10.24963/ijcai.2026/238
UR  - https://doi.org/10.24963/ijcai.2026/238
ER  -