For some ML problems, a bespoke graphical model may fit better than an LLM

roydanroy · x · 2026-07-26

In a short reply, the author suggests that some problems may be better solved by a bespoke graphical model approach rather than an LLM.

They cite a consulting case where, after understanding the domain, the best fit appeared to be an older-style graphical model—possibly an LDA variant—similar to methods that were popular a decade ago. The point is that not every ML problem should be forced into an LLM-shaped solution.

Related event: LLMs Questioned on Tabular Data: Custom Graph Models and PFNs Prove Superior(9 posts)→

Original post →

More from Research

Research channel →