Token-Weighted Multi-Target Learning for Generative Recommenders with Curriculum Learning DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 CCS: Information systems Rec
Wei-Ning Chiu, Chuan-Ju Wang, Pu-Jen Cheng · 2026
Read Token-Weighted Multi-Target Learning for Generative Recommenders with Curriculum Learning DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 CCS: Information systems Rec by Wei-Ning Chiu, Chuan-Ju Wang, Pu-Jen Cheng (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-2601-17787.txt.
Description
Wei-Ning Chiu, Chuan-Ju Wang, and Pu-Jen Cheng work on teaching recommender systems what matters most in a user history. Their 2026 preprint is about generative recommenders. These systems treat the task of picking the next item as a step-by-step text task. Items are named by semantic codes, and the model writes the code of the item it predicts.
Most training treats each piece of the code as worth the same. The authors say that is a poor match for how the codes carry meaning. They propose two ways to weight the pieces by the information each one adds. A staged training plan, called curriculum learning, supports the scheme. The aim is a model that learns the key signals first.
Catalog record
| CREATOR | Wei-Ning Chiu, Chuan-Ju Wang, Pu-Jen Cheng |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-17787