Source record

Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2607.17948.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

Research paper examining whether LLM-driven decisions can replace or enrich rule-based agents in agent-based models. Using Mesa and the Schelling segregation model, one hybrid agent delegates neighbor classification to an LLM via tool calls. Smaller models fail semantic checks; larger models pass. Statistical model checking evaluates simulation reliability.

Empirical test of LLM-agent reliability and tool-call stability inside a standard ABM, with concrete thresholds on model size and semantic accuracy.

Cited by

No citations on record.