# Natural Language Understanding on GLUE (Full Suite)

Source: [Natural Language Understanding on GLUE (Full Suite)](<https://www.sota2.com/research/sota/natural-language-understanding-on-glue-full-suite>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:39:26.981Z

Evaluation of general language understanding capabilities across the full set of constituent tasks in the GLUE benchmark suite (CoLA, SST-2, MRPC, QNLI, QQP, RTE, MNLI-m, MNLI-mm).

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Natural Language Understanding |
| Dataset (including any stated split) | GLUE |
| Metric | CoLA Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [CoLA Score](<https://www.sota2.com/research/sota/natural-language-understanding-on-glue-full-suite.md?metric=CoLA%20Score>): Higher is better; unit: PERCENT.

- [MNLI-m Acc](<https://www.sota2.com/research/sota/natural-language-understanding-on-glue-full-suite.md?metric=MNLI-m%20Acc>): Higher is better; unit: PERCENT.

- [MNLI-mm Acc](<https://www.sota2.com/research/sota/natural-language-understanding-on-glue-full-suite.md?metric=MNLI-mm%20Acc>): Higher is better; unit: PERCENT.

- [MRPC Acc](<https://www.sota2.com/research/sota/natural-language-understanding-on-glue-full-suite.md?metric=MRPC%20Acc>): Higher is better; unit: PERCENT.

- [QNLI Acc](<https://www.sota2.com/research/sota/natural-language-understanding-on-glue-full-suite.md?metric=QNLI%20Acc>): Higher is better; unit: PERCENT.

- [QQP Acc](<https://www.sota2.com/research/sota/natural-language-understanding-on-glue-full-suite.md?metric=QQP%20Acc>): Higher is better; unit: PERCENT.

- [RTE Acc](<https://www.sota2.com/research/sota/natural-language-understanding-on-glue-full-suite.md?metric=RTE%20Acc>): Higher is better; unit: PERCENT.

- [SST-2 Acc](<https://www.sota2.com/research/sota/natural-language-understanding-on-glue-full-suite.md?metric=SST-2%20Acc>): Higher is better; unit: PERCENT.

## Recorded results

| Method | CoLA Score | Result date | Report paper |
| --- | --- | --- | --- |
| ACL | 63.1 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| RoBERTa (CE+SCL) | 62.7 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| RoBERTa (CE) | 62.4 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| ACL | 58.5 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| BERT (CE+SCL) | 57.9 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| BERT (CE) | 57.3 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| TinyBERT (6L) | 46.3 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| ACL-CL | 46.2 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| DistilBERT (6L) | 46.1 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| HeadPrune | 45.7 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| RoBERTa-6L | 45.6 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| BERT-of-Theseus | 44.7 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| ACL-CL | 44.7 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| LayerDrop-6L | 43.7 | Not recorded | 6982d2b6a8b01e76cc8e965f |
| BERT-6L | 43.2 | Not recorded | 6982d2b6a8b01e76cc8e965f |

### ACL

CoLA Score: 63.1. Result ID: 69964f705698665ec494247e.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

Recorded evaluation configuration: {"Model":"RoBERTa","Method":"ACL","#Flops":"6.7G"}. This record does not assert independent reproduction.

### RoBERTa (CE+SCL)

CoLA Score: 62.7. Result ID: 69964f705698665ec494247d.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69b41e209131cea2393a.md>)

Recorded evaluation configuration: {"Model":"RoBERTa","Method":"CE+SCL","#Flops":"6.7G"}. This record does not assert independent reproduction.

### RoBERTa (CE)

CoLA Score: 62.4. Result ID: 69964f705698665ec494247c.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66131e209131cea0e18b.md>)

[Implementation](<https://github.com/pytorch/fairseq>)

Recorded evaluation configuration: {"Model":"RoBERTa","Method":"CE","#Flops":"6.7G"}. This record does not assert independent reproduction.

### ACL

CoLA Score: 58.5. Result ID: 69964f705698665ec494247b.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

Recorded evaluation configuration: {"Model":"BERT","Method":"ACL","#Flops":"6.6G"}. This record does not assert independent reproduction.

### BERT (CE+SCL)

CoLA Score: 57.9. Result ID: 69964f705698665ec494247a.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69b41e209131cea2393a.md>)

Recorded evaluation configuration: {"Model":"BERT","Method":"CE+SCL","#Flops":"6.6G"}. This record does not assert independent reproduction.

### BERT (CE)

CoLA Score: 57.3. Result ID: 69964f705698665ec4942479.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65c31e209131cea0c4ed.md>)

[Implementation](<https://github.com/google-research/bert>)

Recorded evaluation configuration: {"Model":"BERT","Method":"CE","#Flops":"6.6G"}. This record does not assert independent reproduction.

### TinyBERT (6L)

CoLA Score: 46.3. Result ID: 69964f705698665ec4942484.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa688a1e209131cea1cce4.md>)

[Implementation](<https://github.com/huawei-noah/Pretrained-Language-Model/tree/master/TinyBERT>)

Recorded evaluation configuration: {"Model":"TinyBERT (6L)","Method":"-","#Flops":"3.3G"}. This record does not assert independent reproduction.

### ACL-CL

CoLA Score: 46.2. Result ID: 69964f705698665ec4942487.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

Recorded evaluation configuration: {"Model":"ACL-RoBERTa-6L","Method":"ACL-CL","#Flops":"3.3G"}. This record does not assert independent reproduction.

### DistilBERT (6L)

CoLA Score: 46.1. Result ID: 69964f705698665ec4942483.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67221e209131cea14884.md>)

[Implementation](<https://github.com/huggingface/transformers>)

Recorded evaluation configuration: {"Model":"DistilBERT (6L)","Method":"-","#Flops":"3.3G"}. This record does not assert independent reproduction.

### HeadPrune

CoLA Score: 45.7. Result ID: 69964f705698665ec4942482.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa669b1e209131cea1157c.md>)

Recorded evaluation configuration: {"Model":"HeadPrune","Method":"-","#Flops":"4.7G"}. This record does not assert independent reproduction.

### RoBERTa-6L

CoLA Score: 45.6. Result ID: 69964f705698665ec4942480.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66131e209131cea0e18b.md>)

[Implementation](<https://github.com/pytorch/fairseq>)

Recorded evaluation configuration: {"Model":"RoBERTa-6L","Method":"-","#Flops":"3.4G"}. This record does not assert independent reproduction.

### BERT-of-Theseus

CoLA Score: 44.7. Result ID: 69964f705698665ec4942485.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa686f1e209131cea1c320.md>)

Recorded evaluation configuration: {"Model":"BERT-of-Theseus","Method":"-","#Flops":"3.3G"}. This record does not assert independent reproduction.

### ACL-CL

CoLA Score: 44.7. Result ID: 69964f705698665ec4942486.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

Recorded evaluation configuration: {"Model":"ACL-BERT-6L","Method":"ACL-CL","#Flops":"3.3G"}. This record does not assert independent reproduction.

### LayerDrop-6L

CoLA Score: 43.7. Result ID: 69964f705698665ec4942481.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66771e209131cea104ba.md>)

Recorded evaluation configuration: {"Model":"LayerDrop-6L","Method":"-","#Flops":"3.3G"}. This record does not assert independent reproduction.

### BERT-6L

CoLA Score: 43.2. Result ID: 69964f705698665ec494247f.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2b6a8b01e76cc8e965f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65c31e209131cea0c4ed.md>)

[Implementation](<https://github.com/google-research/bert>)

Recorded evaluation configuration: {"Model":"BERT-6L","Method":"-","#Flops":"3.3G"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
