Multi-label

DataFunTalk

Mar 18, 2023 · Artificial Intelligence

Evaluating ChatGPT on Complex Knowledge‑Base Question Answering Using a Feature‑Driven Multi‑Label Framework

This study presents a comprehensive evaluation of ChatGPT's ability to answer complex knowledge‑base questions by introducing a feature‑driven multi‑label classification framework and applying the CheckList black‑box testing methodology across eight KB‑based CQA datasets, comparing its performance with GPT‑3, GPT‑3.5 and FLAN‑T5.