akoksal/LongForm
Viewer
•
Updated
•
27.7k
•
217
•
57
We propose the reverse instruction method to create instruction tuning dataset from corpus examples and LLMs. Our dataset and set of finetuned models:
Note The LongForm dataset.
Note LongForm-LLaMA model. We only share the diff, you can get the final model by downloading the base model and following the code example in our model doc.
Note LongForm-T5 model with the encoder-decoder architecture.
Note The smallest LongForm model. It is capable of long-form text generation and following instructions.