kt-ai-challenge

개요

https://aifactory.space/task/2528/overview
[결선] 2023 AI(인공지능) 대학원 챌린지 with kt 믿:음 / xfact팀

주제

• 설명가능 하고, 개인 맞춤 가능한 멀티 홉 질의응답 (Explainable and Customizable Multi-hop Question Answering)

Demo

Video

https://youtu.be/K9_glUcdj1E

Folder Descriptions

data

• data directory contains datasets used for training/inference on KT'S 믿:음
• For further details and descriptions, check out README.md in data directory

frontend

• frontend directory contains scripts and codes used for Demo.
• For further details and descriptions, check out README.md in frontend directory

src

• src directory contains utility functions for calling apis/files.

Script Descriptions

• kt-evaluate-api.py : evaluation script for default/trained KT믿:음 models on reasoning dataset
• reasoning_api.py : reasoning script for adding rationales behind the answer using LLM apis
• translate_api.py : translation script from English to Korean using LLM apis

License

Apache 2.0 license