LmRaC

LmRaC interrogates user experimental results by building domain-specific knowledge bases from PubMed sources and supporting quantitative experimental data through a customizable REST API.

Key Features:

• RAGdom: Dynamically builds domain-specific knowledge bases from PubMed sources with paragraph-level citations.
• RAGexp: Constructs experimental context from answers drawn from RAGdom, along with user-supplied documents such as design and protocols.
• RAGfun: Supports questions about quantitative experimental data through a user-defined and functionally extensible REST API server.
• Hallucination elimination: Answers are drawn solely from the retrieval-augmented generation knowledge base with citations to the paragraph level, virtually eliminating any chance of hallucination or fabrication.

Scientific Applications:

• Experimental result interpretation: Answers complex scientific questions in the context of a user's own experimental results.
• Quantitative data analysis: Integrates quantitative experimental results to answer questions about specific experiments.
• Controlled source tracing: Ensures results are accurate, complete, and authoritative by carefully controlling and tracing source material.

Methodology:

LmRaC uses retrieval-augmented generation to make data and documents available to the large language model for reasoning tasks.