G 2 LoRA: Gradient Orthogonal Low-Rank Adaptation Framework for Graph Continual Learning on Text-Attributed Graphs Conference: Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2; August 09–13, 2026; Jeju Island, Republic of Korea Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD ’26), August 09–13, 2026, Jeju Island, Republic of Korea DOI: 10.1145/3770855.3817966 ISBN: 979-8-4007-2259-2/2026/08 CCS: Information systems Data mining
Yuhan Wang, Yibo Ding, Yutong Ye, Mufan Zhao, Wenbo Zhang, Ruijie Wang, Jianxin Li · 2026
Read G 2 LoRA: Gradient Orthogonal Low-Rank Adaptation Framework for Graph Continual Learning on Text-Attributed Graphs Conference: Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2; August 09–13, 2026; Jeju Island, Republic of Korea Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD ’26), August 09–13, 2026, Jeju Island, Republic of Korea DOI: 10.1145/3770855.3817966 ISBN: 979-8-4007-2259-2/2026/08 CCS: Information systems Data mining by Yuhan Wang, Yibo Ding, Yutong Ye, Mufan Zhao, Wenbo Zhang, Ruijie Wang, Jianxin 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-2606-01873.txt.
Description
Wang and co-authors study graph models that must learn tasks in a stream. The graphs carry text on their nodes. A pre-training scheme aligns graph and text codes. It works well on single tasks done alone. But tasks fed in sequence cause sharp forgetting.
Light tuning methods ease forgetting only in part. Task interference and weak transfer remain as core flaws. Mixed tasks pull the training aims in varied ways. Graph and text encoders also react at varied rates. Their G2LoRA method uses gradient-orthogonal low-rank updates. The aim is to cut interference between tasks. It also seeks to boost positive transfer across tasks. Tests probe how well the balance holds.
Catalog record
| CREATOR | Yuhan Wang, Yibo Ding, Yutong Ye, Mufan Zhao, Wenbo Zhang, Ruijie Wang, Jianxin Li |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.01873