Fine-Grained Graph Generation through Latent Mixture Scheduling
Nidhi Vakil, Hadi Amiri · 2026
Read Fine-Grained Graph Generation through Latent Mixture Scheduling by Nidhi Vakil, Hadi Amiri (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-2605-02780.txt.
Description
Nidhi Vakil and Hadi Amiri build a graph generator that can be steered in fine detail, not just in broad strokes. Generating graphs with set structural traits serves drug discovery, social network modeling, and knowledge graph work. Prior methods offer only coarse control over those traits, which limits their use when exact structure matters.
Their model is a conditional variational autoencoder whose decoder space is refined during generation. A mixture scheduler gradually blends graph-driven and property-driven signals, aligning the two as a graph takes shape. That dynamic alignment aims to raise both realism and obedience to the requested traits. On five real-world datasets, the model produced high-quality graphs while keeping strong control, beating recent baselines. Fine control, the authors argue, is what makes generated graphs usable.
Catalog record
| CREATOR | Nidhi Vakil, Hadi Amiri |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.02780