search

ZHZisZZ / weak-to-strong-search

[NeurIPS'24] Weak-to-Strong Search: Align Large Language Models via Searching over Small Language Models
https://arxiv.org/abs/2405.19262
39 stars 3 forks source link

dataset ZHZisZZ/imdb_preference #3

Closed Young-Zhen closed 1 week ago

Young-Zhen commented 1 week ago

Hi, This is actually a remarkable work in the fild of LLM alignment, thanks for your marvelous work! But I got some problems when reimplementing the results of controlled sentiment generation, that is, the dataset 'ZHZisZZ/imdb_preference' is missing so the fine-tuning of base gpt-2 models is not easy to carry out. Therefore, I wonder if you mind sharing the dataset with a google drive link or baidu netdisk link? I believe it will, to some extent, help me and other people to follow this work. Finally, this is absolutely a pioneering work in my view. Thanks for your excellent work and the contributions to the open-source comunity!

ZHZisZZ commented 1 week ago

This dataset is actually publically available: https://huggingface.co/datasets/ZHZisZZ/imdb_preference.

If you run the training script, the dataset will be automatically downloaded.

Young-Zhen commented 1 week ago

sorry, I have found the dataset in hugging face right now. very sorry to bother you!