Rethinking Contrastive Learning for Graph Collaborative Filtering: Limitations and a Simple Remedy
Geon Lee, Sunwoo Kim, Kyungho Kim, Kijung Shin · 2026
Read Rethinking Contrastive Learning for Graph Collaborative Filtering: Limitations and a Simple Remedy by Geon Lee, Sunwoo Kim, Kyungho Kim, Kijung Shin (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-2605-24015.txt.
Description
Lee and three co-authors unpack how contrastive learning interacts with graph collaborative filtering, where scores aggregate learnable weights over multi-hop neighbour pairs. They show good recommendations can come from upweighting only a small subset of pairs whose neighbours resemble the target user and item, with effects varying by pair type. Analyzing the Sampled Softmax loss, they find its weight-update dynamics flawed in key ways. Their fix, NT-SSM, induces type-aware update dynamics and beats Sampled Softmax consistently across data sets and filtering models.
Catalog record
| CREATOR | Geon Lee, Sunwoo Kim, Kyungho Kim, Kijung Shin |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 75402 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.24015