Source record

Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2609.01167.

Pipeline notes

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

Paper presents EpisodeSim, a hybrid LLM-agent architecture that encodes classic-AI structures—roles, scripts, scenes, obligations, closure—as natural-language control state interpreted by LLM calls. A World Master maintains shared reality and adjudicates actions. Qualitative ablations on two settings argue LLM fluency alone is insufficient for coherent social simulation.

Concrete architecture pattern for role-based social agents: World Master, scene construction, obligation tracking. Useful blueprint for anyone building bounded multi-turn simulations.

Cited by

No citations on record.