TALON: Test-time Adaptive Learning for On-the-Fly Category Discovery
Yanan Wu, Yuhan Yan, Tailai Chen, Zhixiang Chi, ZiZhang Wu, Yi Jin, Yang Wang, Zhenbo Li · 2026
Read TALON: Test-time Adaptive Learning for On-the-Fly Category Discovery by Yanan Wu, Yuhan Yan, Tailai Chen, Zhixiang Chi, ZiZhang Wu, Yi Jin, Yang Wang, Zhenbo Li (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-2603-08075.txt.
Description
Wu and colleagues propose TALON, a method for on-the-fly category discovery. A model must recognize known categories while also spotting novel ones in an unlabeled online stream, having trained only on labeled data. Existing approaches freeze the offline-trained feature extractor and quantize features into binary codes as class prototypes. But a fixed knowledge base ignores what incoming data could teach. Quantization also loses information, weakens expressiveness, and amplifies variance within classes. TALON instead adapts its learning at test time as new data arrives.
Catalog record
| CREATOR | Yanan Wu, Yuhan Yan, Tailai Chen, Zhixiang Chi, ZiZhang Wu, Yi Jin, Yang Wang, Zhenbo Li |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45247 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.08075