Reproduce “Kierston & KierstonBench: Simulation of Allentown, PA and Evaluation of Language Models in Crime Investigation and Surveillance Tasks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Andrew Semenov, Svyatoslav Dorofeev
Kierston simulates city residents’ activities and derives provider records from a shared timeline, which models query through a restricted interface without seeing activity histories or incident roles. KierstonBench scores conclusions against simulated ground truth across 32 tasks. GPT-6 Sol averaged 0.55 on the −1-to-1 task scale, but no model identified a perpetrator in five suspect-identification tasks in this single-attempt evaluation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.