As a RAG task, why wasn't the context content used during evaluation (batch_eval)? If there are possibilities of hallucinations, this evaluation might not be quite accurate. - Githubissues

HKUDS / LightRAG

"LightRAG: Simple and Fast Retrieval-Augmented Generation"

https://arxiv.org/abs/2410.05779

MIT License

9.26k stars 1.14k forks source link

As a RAG task, why wasn't the context content used during evaluation (batch_eval)? If there are possibilities of hallucinations, this evaluation might not be quite accurate. #251

Open luoling1993 opened 1 week ago

luoling1993 commented 1 week ago

As the title.