Xingcheng Yao

Hi~ Thanks for your interest in our work. The difference between scales in TLM is essentially the hyper-parameters we choose that affect the training FLOPs (such as batch size, training...

You are welcome. But we do use **test marco_f1** as the reported metric in the main results. The point is we don't use test set results evaluated in the final...

We haven't found a publicly available version of STORIES yet. In our work, we follow the same methodology in [the original paper of STORIES dataset](https://arxiv.org/pdf/1806.02847.pdf) (Section 5.3) to collect a...

Hi. I'm sorry have no clear recollection about this error in my experience. My suggestions are 1. Make sure elasticsearch is really started after executing `./bin/elasticsearch` by `curl localhost:9200`. 2....

Hello~ The main function of the code is in `src/run.py` starting from line 435. You may follow the instructions in `README.md` to run the code.

Oh, I'm so sorry. Due to the lack of resources both in machines and in human, the data retrieval service is no longer maintained. I've already modified that part in...

Sorry for very late reply (I've been busy since last year), hope I could still be of help. huggingface_hub got rid of hf_api long ago, so updating wouldn't work. I...