RESEARCH · RESEARCH · #2170
Study: graph serialization choice substantially affects LLM structural reasoning
The paper (arXiv:2610.22734v1) evaluates 15 graph serialization formats across nine frontier LLM configurations on ontology-based knowledge extraction and structural graph reasoning tasks. It finds knowledge extraction is relatively insensitive to format, but structural reasoning accuracy depends strongly on the serialization, with the best format varying by model, query, and task; the authors propose treating serialization as a dynamic orchestration decision.
KEY POINTS
- The paper (arXiv:2610.22734v1) evaluates 15 graph serialization formats across nine frontier LLM configurations on ontology-based knowledge extraction and structural graph reasoning tasks.
- It finds knowledge extraction is relatively insensitive to format, but structural reasoning accuracy depends strongly on the serialization, with the best format varying by model, query, and task; the authors propose treating serialization as a dynamic orchestration decision.
- The result matters because it shows how the textual encoding of knowledge graphs can materially change LLM reasoning performance and suggests systems should select or adapt formats per model and task.
WHY IT MATTERS
The result matters because it shows how the textual encoding of knowledge graphs can materially change LLM reasoning performance and suggests systems should select or adapt formats per model and task.