18 datasets found

The BIG-Bench Hard dataset is derived from the original BIG-Bench evaluation suite, focusing on tasks that pose challenges to existing language models.

LongPile is a diverse dataset derived from the Pile corpus.

The TruthfulQA dataset is a dataset that contains 817 questions designed to evaluate language models' preference to mimic some human falsehoods.

The paper proposes a novel approach to aligning language models with human preferences, focusing on the use of preference optimization in reward-free RLHF.

The LLaMA dataset is a large language model dataset used in the paper.

The dataset used in the paper for instruction following task

HONEST is a fairness dataset specifically designed to assess LMs' outputs' hurtfulness.

FAIRBELIEF is a language-agnostic analytical approach to capture and assess beliefs embedded in LMs.

The GMEG-wiki and GMEG-yahoo datasets are used to evaluate the proposed approach.

The Break-It-Fix-It (BIFI) framework has demonstrated strong results on learning to repair a broken program without any labeled examples.

The task of grammatical error correction (GEC) is to map an ungrammatical sentence xbad into a grammatical version of it, xgood.

Training a model for grammatical error correction (GEC) requires a set of labeled ungrammatical / grammatical sentence pairs, but manually annotating such pairs can be expensive.

The dataset used in the paper is not explicitly described, but it is mentioned that the authors used three downstream tasks from the GLUE benchmark: Stanford Sentiment Treebank...

Human speech data comprises a rich set of domain factors such as accent, syntactic and semantic variety, or acoustic environment.

This paper proposes BERT, a pre-trained deep bidirectional transformer for language understanding.

The DailyDialog dataset is a large-scale multi-turn dialogue dataset, consisting of 10,000 conversations with 5 turns each.

Pre-trained language models (PrLM) have to carefully manage input units when training on a very large text with a vocabulary consisting of millions of words. Previous works have...

The dataset used in the paper is not explicitly described, but it is mentioned that the authors used a condensed representation of LLM activations obtained from sparse...