Roberta Large Wanli

🚀 預訓練模型：roberta-large-wanli

這是一個基於Worker-AI協作自然語言推理數據集WANLI微調的roberta-large現成模型（Liu等人，2022）。在八個域外測試集上，它的表現優於roberta-large-mnli模型，在HANS上提升了11%，在對抗性自然語言推理數據集上提升了9%。

🚀 快速開始

模型信息

示例輸入

• "I almost forgot to eat lunch.I didn't forget to eat lunch."
• "I almost forgot to eat lunch.I forgot to eat lunch."
• "I ate lunch.I almost forgot to eat lunch."

💻 使用示例

基礎用法

from transformers import RobertaTokenizer, RobertaForSequenceClassification

model = RobertaForSequenceClassification.from_pretrained('alisawuffles/roberta-large-wanli')
tokenizer = RobertaTokenizer.from_pretrained('alisawuffles/roberta-large-wanli')

x = tokenizer("I almost forgot to eat lunch.", "I didn't forget to eat lunch.", return_tensors='pt', max_length=128, truncation=True)
logits = model(**x).logits
probs = logits.softmax(dim=1).squeeze(0)
label_id = torch.argmax(probs).item()
prediction = model.config.id2label[label_id]

📚 詳細文檔

引用信息

@inproceedings{liu-etal-2022-wanli,
    title = "{WANLI}: Worker and {AI} Collaboration for Natural Language Inference Dataset Creation",
    author = "Liu, Alisa  and
      Swayamdipta, Swabha  and
      Smith, Noah A.  and
      Choi, Yejin",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2022",
    month = dec,
    year = "2022",
    address = "Abu Dhabi, United Arab Emirates",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2022.findings-emnlp.508",
    pages = "6826--6847",
    abstract = "A recurring challenge of crowdsourcing NLP datasets at scale is that human writers often rely on repetitive patterns when crafting examples, leading to a lack of linguistic diversity. We introduce a novel approach for dataset creation based on worker and AI collaboration, which brings together the generative strength of language models and the evaluative strength of humans. Starting with an existing dataset, MultiNLI for natural language inference (NLI), our approach uses dataset cartography to automatically identify examples that demonstrate challenging reasoning patterns, and instructs GPT-3 to compose new examples with similar patterns. Machine generated examples are then automatically filtered, and finally revised and labeled by human crowdworkers. The resulting dataset, WANLI, consists of 107,885 NLI examples and presents unique empirical strengths over existing NLI datasets. Remarkably, training a model on WANLI improves performance on eight out-of-domain test sets we consider, including by 11{\%} on HANS and 9{\%} on Adversarial NLI, compared to training on the 4x larger MultiNLI. Moreover, it continues to be more effective than MultiNLI augmented with other NLI datasets. Our results demonstrate the promise of leveraging natural language generation techniques and re-imagining the role of humans in the dataset creation process.",
}