Researchers develop new way to reduce AI hallucinations
Researchers develop new way to reduce AI hallucinations
Publish Date: 2026-05-28 03:00:00
Source Domain: www.binghamton.edu
- The rise of AI-powered chatbots has led to increased use in diagnosing medical concerns, highlighting the need for accurate responses.
- Binghamton University researchers tested Open AI’s ChatGPT, finding it accurate but prone to “hallucinations,” or falsely generated information.
- A new verification method, co-developed by Ahmed Abdeen Hamed and Luis M. Rocha, uses a consensus voting system with seven large language models to reduce hallucinations.
- The new protocol employs retrieval-augmented generation (RAG) to reference authoritative databases for verifying medical terminology and diagnoses.
- The verification method can be adapted to various contexts, increasing confidence in large multiscale network models of disease and allowing for precision medicine applications.
