self-taught
self-taught-reasoner-rollouts
STaR Rationale Dataset (CommonsenseQA, Llama-2-7B)
This repository contains bootstrapped rationale datasets produced by a small replication of STaR (Self-Taught Reasoner) on the CommonsenseQA training split using Llama-2-7B as the base model (M_0).
Each .jsonl file corresponds to one iteration of the STaR pipeline and stores a set of question–rationale–answer triples collected during that iteration.
Data generation procedure (STaR iteration)
Let the training… See the full description on the dataset page: https://huggingface.co/datasets/parksoojae/self-taught-reasoner-rollouts.WIP-Dataset-For-Self-Taught-Evaluators-v2SFT-Dataset-For-Self-Taught-Evaluators-iter1SFT-Dataset-For-Self-Taught-Evaluators-v2-iter1SFT-Dataset-For-Self-Taught-Evaluators-iter2Self-taught-evaluator-DPO-dataThis dataset is released as part of Self-taught evaluators research project.
Please refer to our project materials here for training and evaluation details.
Loading the dataset with transformers
This dataset is built upon WildChat prompts by using Llama-3.1-70B-Instruct to generate responses and evaluation plans. Details on how to build such a self-taught dataset can be found in Self-taught evaluators.
Minimal example below showing how to prepare training data.
from datasets import… See the full description on the dataset page: https://huggingface.co/datasets/facebook/Self-taught-evaluator-DPO-data.