2026 Request for Proposals: Evaluating LLMs for Actuarial Perception Tasks for an LLM Benchmark and Re-Evaluation Suite
CAS seeks research to benchmark LLMs on actuarial perception tasks, creating open datasets, evaluation tools, and a public platform for P&C applications.
Funder: Casualty Actuarial Society
Due Dates: September 28, 2026 (Proposals Due)
Funding Amounts: $25,000–$65,000 typical; up to $75,000 maximum per project
Summary: Supports research to evaluate large language models on actuarial perception tasks and develop a benchmark and re-evaluation suite for P&C applications.
Key Information: No overhead costs allowed; all code/data must be published to CAS GitHub.
Description
This funding opportunity, offered by the Casualty Actuarial Society (CAS), seeks proposals for innovative research evaluating large language models (LLMs) on actuarial perception and classification tasks. The goal is to develop a robust, repeatable benchmarking suite and public comparison platform tailored to property and casualty (P&C) actuarial applications. The project will help the actuarial profession objectively assess LLM performance on tasks where answers are measurable and reproducible, such as claims triage, underwriting, fraud detection, and rating plan validation. Deliverables include datasets, evaluation scripts, scoring methodologies, and a public-facing platform, all to be published openly for ongoing use and re-testing as new models emerge.
