VeriTrail

[ntsarb]

Hello,

I read the article on VeriTrail and wondering whether this could become part of the GraphRAG implementation, to detect and prevent hallucinations in the summaries, in a timely manner (e.g. during the indexing process).

VeriTrail: Detecting hallucination and tracing provenance in multi-step AI workflows