-- This is the model checkpoint of GenRead [2], based on the T5-3B and trained on the TriviaQA [1].

-- Hyperparameters: 8 x 80GB A100 GPUs; batch size 16; AdamW; LR 6e-5; best dev at 8500 steps

References:

[1] TriviaQA: A Large Scale Dataset for Reading Comprehension and Question Answering. ACL 2017

[2] Generate rather than Retrieve: Large Language Models are Strong Context Generators. arXiv 2022

We evaluate it on the TriviaQA dataset, the EM score is 71.55.