FedTIP: Communication-Efficient Federated Temporal Prompting for Few-Shot Dynamic Graph Adaptation
Xijun Wu, Xinming Zhang · 2026
Read FedTIP: Communication-Efficient Federated Temporal Prompting for Few-Shot Dynamic Graph Adaptation by Xijun Wu, Xinming Zhang (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/pmc-13605309.txt.
Description
Xijun Wu and Xinming Zhang present FedTIP, a light-touch federated method for changing interaction graphs in a 2026 paper. Interaction streams sit with separate owners, labels are scarce, and sending big model states back and forth is costly. Prior graph prompts assumed one central trainer.
A frozen encoder, trained once on pooled history, anchors the system. Each client tunes only small time-aware prompts on its own labeled events and sends back a compact message. Updates merge under rules weighing coverage and step size. Across three public streams and nine task cells, FedTIP topped every federated rival while sending a fraction of their data volume.
Catalog record
| CREATOR | Xijun Wu, Xinming Zhang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 56 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |