GLUE datasets and statistics. CoLA: Warstadt et al. (2019); MRPC: Dolan and Brockett (2005); SST-2: Socher et al. (2013); STS-B: Cer et al. (2017); QQP: (accessed September 1, 2020) (2017); MNLI: Williams et al. (2018); QNLI is compiled by GLUE’s authors using Rajpurkar et al. (2016). RTE is the concatenation of Dagan et al. (2005); Bar-Haim et al. (2006); Giampiccolo et al. (2007); Bentivogli et al. (2009).
Data . | Task . | |Train| . | |Dev.| . |
---|---|---|---|
CoLA | Acceptability | 8.5K | 1K |
MRPC | Paraphrase | 2.7K | 409 |
QNLI | Entailment | 105K | 5.5K |
RTE | Entailment | 2.5K | 278 |
SST-2 | Sentiment | 67K | 873 |
STS-B | Similarity | 5.8K | 1.5K |
QQP | Paraphrase | 363K | 40K |
MNLI | Entailment | 392K | 9.8K |
Data . | Task . | |Train| . | |Dev.| . |
---|---|---|---|
CoLA | Acceptability | 8.5K | 1K |
MRPC | Paraphrase | 2.7K | 409 |
QNLI | Entailment | 105K | 5.5K |
RTE | Entailment | 2.5K | 278 |
SST-2 | Sentiment | 67K | 873 |
STS-B | Similarity | 5.8K | 1.5K |
QQP | Paraphrase | 363K | 40K |
MNLI | Entailment | 392K | 9.8K |