Scalability Matters: Overcoming Challenges in InstructGLM with Similarity-Degree-Based Sampling
Hyun Lee 1 2, Chris Yi 1, B.D.S. Aritra, Maminur Islam · 2025
Read Scalability Matters: Overcoming Challenges in InstructGLM with Similarity-Degree-Based Sampling by Hyun Lee 1 2, Chris Yi 1, B.D.S. Aritra, Maminur Islam (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-2505-03799.txt.
Description
Hyun Lee and co-authors make language models work on large graphs without a graph network, in work from 2025. Graphs are too big to feed whole into a language model, since token caps cut the input short. Past work leaned on graph networks as helpers, which adds weight and hides the steps.
Their method samples the graph by walks that favor like nodes and busy hubs. That keeps key facts while using far fewer tokens. Tuned by plain instructions, the model then sorts nodes and guesses links on its own. The team shares its code in public and frames the work as a lean path to graph tasks.
Catalog record
| CREATOR | Hyun Lee 1 2, Chris Yi 1, B.D.S. Aritra, Maminur Islam |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-03799