Research map / Large language models / Agents, code and math
Evaluating LLMs for Police Decision-Making: A Framework Based on Police Action Scenarios
Sangyub Lee, Heedou Kim, Hyeoncheol Kim
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use reference answers from police manuals instead of commercial LLM answers | police scenario QA · Police Action Scenarios QA | overall score (1-5): 2.69 (GPT-4) / 2.83 (Claude) / 3.06 (Gemini) -> 3.88 | Improved |
| LLMs versus references on factuality | GPT-4, Gemini, Claude · Police Action Scenarios QA | factuality (1-5): 4.15 -> 2.60 (GPT-4) | Hurt |
| LLM readability versus reference | GPT-4, Gemini, Claude · Police Action Scenarios QA | readability (1-5): 3.80 -> 3.59 / 3.84 / 3.57 | No clear effect |
| add Logical Explanation as a key metric (ablation) | metric selection · expert correlation | Spearman rho: rho 0.203, p=.055 | No clear effect |
| add Logical Correctness as a key metric (ablation) | metric selection · expert correlation | Spearman rho: rho 0.560 (p<.001) | Improved |
| rely on LLM-as-judge instead of police experts | evaluation · expert alignment | Spearman rho: rho 0.315 to 0.560 | Mixed |
| compare Gemini to GPT-4 as police assistant | commercial LLMs · Police Action Scenarios QA | overall score (1-5): 2.69 -> 3.06 | Improved |
Closest papers by meaning
- DeEscalWild: A Real-World Benchmark for Automated De-Escalation Training with SLMs (NeurIPS 2026 · similarity 0.85)
- NYU CTF Bench: A Scalable Open-Source Benchmark Dataset for Evaluating LLMs in Offensive Security (NeurIPS 2024 · similarity 0.83)
- LLMEval: A Preliminary Study on How to Evaluate Large Language Models (AAAI 2024 · similarity 0.82)
- LexEval: A Comprehensive Chinese Legal Benchmark for Evaluating Large Language Models (NeurIPS 2024 · similarity 0.82)
- RocketEval: Efficient automated LLM evaluation via grading checklist (ICLR 2025 · similarity 0.82)
- SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal (ICLR 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
