WebClipper: Efficient Evolution of Web Agents with Graph-based Trajectory Pruning
Junjie Wang, Zequn Xie, Dan Yang, Jie Feng, Yue Shen, Duolin Sun · 2026
Read WebClipper: Efficient Evolution of Web Agents with Graph-based Trajectory Pruning by Junjie Wang, Zequn Xie, Dan Yang, Jie Feng, Yue Shen, Duolin Sun (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-2602-12852.txt.
Description
Web agents that research online often wander through long, looping tool-call trajectories, wasting time on dead ends. Junjie Wang and eleven co-authors propose WebClipper, which compresses those trajectories by treating search as a state graph. It mines a minimal directed acyclic graph that keeps essential reasoning while cutting redundant steps. Training continues on the pruned trajectories, steering the agent toward leaner search habits.
The payoff is concrete: tool-call rounds fall by about 20% while accuracy improves, and a new F-AE score measures the balance between accuracy and efficiency. Experiments confirm that pruned trajectories preserve performance while cutting cost. For deep research systems, the message is that how an agent searches matters as much as what it finds, and shorter, cleaner paths can also be better ones.
Catalog record
| CREATOR | Junjie Wang, Zequn Xie, Dan Yang, Jie Feng, Yue Shen, Duolin Sun |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-12852