shmsw25 / FActScore

A package to evaluate factuality of long-form generation. Original implementation of our EMNLP 2023 paper "FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation"
https://arxiv.org/abs/2305.14251
MIT License
292 stars 43 forks source link

Adding cost estimates for OpenAI API usage #12

Closed martiansideofthemoon closed 1 year ago

martiansideofthemoon commented 1 year ago

@shmsw25 the current implementation assumes no cache exists. Should we adjust our cost estimates based on the cache?