Spotlighter: Revisiting Prompt Tuning from a Representative Mining View

Yutong Gao, Maoyuan Shao, Xinyang Huang, Chuang Zhu, Yu Weng, Xuan Liu, Lijuan Sun, Guoshun Nan


Abstract
CLIP’s success has demonstrated that prompt tuning can achieve robust cross-modal semantic alignment for tasks ranging from open-domain recognition to fine-grained classification. However, redundant or weakly relevant feature components introduce noise and incur unnecessary computational costs. In this work, we propose Spotlighter, a lightweight token-selection framework that simultaneously enhances accuracy and efficiency in prompt tuning. Spotlighter evaluates each visual token’s activation from both sample-wise and semantic-wise perspectives and retains only the top-scoring tokens for downstream prediction. A class-specific semantic memory bank of learned prototypes refines this selection, ensuring semantic representativeness and compensating for discarded features. To further prioritize informative signals, we introduce a two-level ranking mechanism that dynamically weights token–prototype interactions. Across 11 few-shot benchmarks, Spotlighter outperforms CLIP by up to 11.19% in harmonic mean accuracy and achieves up to 0.8K additional FPS, with only 21 extra parameters. These results establish Spotlighter as an effective and scalable baseline for prompt tuning.
Anthology ID:
2025.findings-emnlp.392
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2025
Month:
November
Year:
2025
Address:
Suzhou, China
Editors:
Christos Christodoulopoulos, Tanmoy Chakraborty, Carolyn Rose, Violet Peng
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
7435–7449
Language:
URL:
https://aclanthology.org/2025.findings-emnlp.392/
DOI:
Bibkey:
Cite (ACL):
Yutong Gao, Maoyuan Shao, Xinyang Huang, Chuang Zhu, Yu Weng, Xuan Liu, Lijuan Sun, and Guoshun Nan. 2025. Spotlighter: Revisiting Prompt Tuning from a Representative Mining View. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 7435–7449, Suzhou, China. Association for Computational Linguistics.
Cite (Informal):
Spotlighter: Revisiting Prompt Tuning from a Representative Mining View (Gao et al., Findings 2025)
Copy Citation:
PDF:
https://aclanthology.org/2025.findings-emnlp.392.pdf
Checklist:
 2025.findings-emnlp.392.checklist.pdf