PAND: Prompt-Aware Neighborhood Distillation for Lightweight Fine-Grained Visual Classification
Qiuming Luo, Yuebing Li, Feng Li, Chang Kong · 2026
Read PAND: Prompt-Aware Neighborhood Distillation for Lightweight Fine-Grained Visual Classification by Qiuming Luo, Yuebing Li, Feng Li, Chang Kong (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-07768.txt.
Description
Teaching a small vision model to tell similar bird species or car models apart is hard when the teacher is a huge vision-language model. Qiuming Luo and three co-authors propose PAND, a two-stage distillation method that first calibrates meaning with adaptive prompt-aware anchors, then transfers local decision structure between neighboring examples.
Splitting those jobs beats fixed prompts and global matching: their ResNet-18 student reaches 76.09 percent accuracy on the CUB-200 bird benchmark, beating a strong baseline by 3.4 points, and leads on three other fine-grained benchmarks. The paper gives practitioners a practical recipe for squeezing large-model knowledge into lightweight classifiers.
Catalog record
| CREATOR | Qiuming Luo, Yuebing Li, Feng Li, Chang Kong |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-07768