MAT-Agent: Adaptive Multi-Agent Training Optimization
Jusheng Zhang, Kaitong Cai, Yijia Fan, Ning yuan Liu, Keze Wang[1ex] · 2025
Read MAT-Agent: Adaptive Multi-Agent Training Optimization by Jusheng Zhang, Kaitong Cai, Yijia Fan, Ning yuan Liu, Keze Wang[1ex] (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-2510-17845.txt.
Description
Jusheng Zhang and four co-authors treat the training of an image labeler as a team job done live by software agents. In their MAT-Agent system, agents adjust data changes, the optimizer, learning rates, and loss terms while training runs. A bandit method balances trying new settings against keeping ones that already work well.
Rewards blend accuracy, rare-class results, and steady training. On three public image sets, the system beat the compared methods on the main scores the authors track. The paper gives a concrete design for readers who want training that adapts itself instead of following fixed settings.
Catalog record
| CREATOR | Jusheng Zhang, Kaitong Cai, Yijia Fan, Ning yuan Liu, Keze Wang[1ex] |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 111 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-17845