Bus-Conditioned Zero-Shot Trajectory Generation via Task Arithmetic
Shuai Liu, Ning Cao, Yile Chen, Yue Jiang, Gao Cong · 2026
Read Bus-Conditioned Zero-Shot Trajectory Generation via Task Arithmetic by Shuai Liu, Ning Cao, Yile Chen, Yue Jiang, Gao Cong (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-2602-13071.txt.
Description
Mobility data power smart-city planning, but many cities cannot share real trajectories, leaving generation methods stranded. Shuai Liu and four co-authors pose a new setting: trajectory generation conditioned on bus data, with zero target-city trajectories available. Generation uses only source-city mobility data plus public bus timetables from both cities.
Their method, MobTA, brings task arithmetic into trajectory generation for the first time. It models the parameter shift from timetable-based to mobility-based generation in the source city, then applies that shift to the target city through arithmetic on task vectors. Experiments show MobTA beating existing methods and approaching models fine-tuned on target-city data. Stability is also analysed theoretically across base and instruction-tuned language models.
Catalog record
| CREATOR | Shuai Liu, Ning Cao, Yile Chen, Yue Jiang, Gao Cong |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 96 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-13071