search

SqueezeAILab / LLM2LLM

[ACL 2024] LLM2LLM: Boosting LLMs with Novel Iterative Data Enhancement
https://arxiv.org/abs/2403.15042
MIT License
149 stars 10 forks source link

What is instruction_per_seed_task in your experiment? #2

Closed Tokgoleo closed 6 months ago

Tokgoleo commented 6 months ago

It seems that instruction_per_seed_task is not reported in the Experiment Setup section of your paper? I find in your code that it's four by default.

dragon18456 commented 6 months ago

For LLM2LLM, this value is 1. It is set here in the config: https://github.com/SqueezeAILab/LLM2LLM/blob/main/GSM8K/config.yaml#L8 You can use it to set the total number of augmentations per seed task.