2026 Request for Proposals: Evaluating LLMs for Actuarial Perception Tasks for an LLM Benchmark and Re-Evaluation Suite
CAS seeks research to develop a benchmark framework for evaluating large language models on key actuarial perception tasks in property and casualty insurance.
Funder: Casualty Actuarial Society
Due Dates: September 28, 2026: Proposal Deadline
Funding Amounts: $25,000–$65,000 typical award; maximum budget $75,000; project duration up to ~9 months
Summary: Supports research to develop and deliver a benchmarking framework for evaluating large language models on actuarial perception tasks in property and casualty insurance.
Description
This opportunity, offered by the Casualty Actuarial Society (CAS) Artificial Intelligence Working Group, seeks proposals for research projects that will design, implement, and deliver a robust benchmarking framework to evaluate large language models (LLMs) on perception-focused tasks relevant to property and casualty (P&C) actuarial practice. The initiative aims to establish a comprehensive, versioned benchmark and re-evaluation suite—including datasets, evaluation protocols, and a public comparison platform—to objectively assess LLM performance on tasks such as claims triage, underwriting decisions, fraud flagging, and other classification or recognition challenges central to actuarial work. The resulting suite should enable ongoing, reproducible evaluation as new LLMs are released and should be extensible for future research and industry needs.
