Exploiting Task Relationships for Continual Learning Using Transferability-Aware Task Embeddings
Yanru Wu, Xiangyu Chen, Jianning Wang, Enming Zhang, Hanbing Liu, Yang Li · 2025
Read Exploiting Task Relationships for Continual Learning Using Transferability-Aware Task Embeddings by Yanru Wu, Xiangyu Chen, Jianning Wang, Enming Zhang, Hanbing Liu, Yang Li (2025) 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-2502-11609.txt.
Description
Yanru Wu and six co-authors help continual learning systems use the kinship between tasks instead of treating each task as a stranger. Their paper appeared in two thousand twenty-five. Continual learning adds new tasks over time without forgetting old ones. Most methods guard the weights but ignore how tasks relate.
Their task fingerprint, the H-embedding, scores how well skill moves between tasks. It comes from an information measure and is cheap to compute on the fly. A hypernetwork then reads the fingerprint and builds weights fitted to each task. Only the small fingerprint must be stored per task. Training runs end to end without fuss. On image benches the method beat strong rivals at both forward and backward transfer. The code is shared in public.
Catalog record
| CREATOR | Yanru Wu, Xiangyu Chen, Jianning Wang, Enming Zhang, Hanbing Liu, Yang Li |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-11609