Intelligent

ReaLLM adopts the latest LLMs from the market. From math, coding, to medical diagnosis and legal document processing, it's as capable as the best.

Reliable

A deep integration with automated reasoning techniques ensures that the responses of LLMs are logically consistent and trustworthy. It's not about accuracy -- ReaLLM doesn't go wrong.

Overview

ReaLLM integrates a variety of formal methods and automated reasoning techniques, including program refinement, verification, first-order logic solving, with LLMs at different levels. This integration and interaction between the two technologies guides the LLM into making correct responses. Like the simplified view below.

overview

In math reasoning benchmarks, it correctly solves 11% more problems than GPT4 + CoT, and more importantly, it doesn't give wrong answers!

Benchmark

See it live:

ReaLLM for Math

ReaLLM for Health

Conference Paper.

Applications

Learning & Teaching

ReaLLM for Education is a robust tool for generating questions and answers, as well as automatically marking complex exams and assignments.

AI Coding

ReaLLM for Coding enhances LLMs with correct-by-construction coding. Every piece of output code is mathematically proven correct!

AI for Health

ReaLLM for Health helps medical practitioners with diagnosis and analysis with high confidence.

AI for Law and Regulation

ReaLLM for Law reasons about law cases, legal documents and contracts, and provides verdicts and summarise for professionals.