# Natural Language Understanding on SuperGLUE 1,000 examples

Source: [Natural Language Understanding on SuperGLUE 1,000 examples](<https://www.sota2.com/research/sota/natural-language-understanding-on-superglue-1-000-examples>)

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

Data updated: 2026-02-26T05:37:53.797Z

Evaluation of general language understanding capabilities using a 1,000-example subset of the SuperGLUE benchmark suite.

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) | SuperGLUE 1,000 examples |
| Metric | BoolQ Accuracy |
| 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

- [BoolQ Accuracy](<https://www.sota2.com/research/sota/natural-language-understanding-on-superglue-1-000-examples.md?metric=BoolQ%20Accuracy>): Higher is better; unit: PERCENT.

- [RTE Accuracy](<https://www.sota2.com/research/sota/natural-language-understanding-on-superglue-1-000-examples.md?metric=RTE%20Accuracy>): Higher is better; unit: PERCENT.

- [WIC Accuracy](<https://www.sota2.com/research/sota/natural-language-understanding-on-superglue-1-000-examples.md?metric=WIC%20Accuracy>): Higher is better; unit: PERCENT.

- [SST-2 Accuracy](<https://www.sota2.com/research/sota/natural-language-understanding-on-superglue-1-000-examples.md?metric=SST-2%20Accuracy>): Higher is better; unit: PERCENT.

- [Average Accuracy](<https://www.sota2.com/research/sota/natural-language-understanding-on-superglue-1-000-examples.md?metric=Average%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | BoolQ Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| LoRA | 84 | Not recorded | 68fa725a1e209131cea55e46 |
| S-MeZO | 82.2 | Not recorded | 68fa725a1e209131cea55e46 |
| MeZO-LoRA | 80.3 | Not recorded | 68fa725a1e209131cea55e46 |
| AdaZeta | 79.8 | Not recorded | 68fa725a1e209131cea55e46 |
| ZO-SGD-Adam | 78.9 | Not recorded | 68fa725a1e209131cea55e46 |
| MeZO | 78.8 | Not recorded | 68fa725a1e209131cea55e46 |
| ZO-AdaMU | 78.2 | Not recorded | 68fa725a1e209131cea55e46 |
| ZO-SGD-Cons | 77.1 | Not recorded | 68fa725a1e209131cea55e46 |
| R-MeZO | 76.7 | Not recorded | 68fa725a1e209131cea55e46 |
| S-MeZO | 73.8 | Not recorded | 68fa725a1e209131cea55e46 |
| R-MeZO | 72.3 | Not recorded | 68fa725a1e209131cea55e46 |
| MeZO | 72.1 | Not recorded | 68fa725a1e209131cea55e46 |
| ZO-SGD-Sign | 68.2 | Not recorded | 68fa725a1e209131cea55e46 |
| ICL | 66.9 | Not recorded | 68fa725a1e209131cea55e46 |
| Zero Shot | 59 | Not recorded | 68fa725a1e209131cea55e46 |

### LoRA

BoolQ Accuracy: 84. Result ID: 699a9efd5698665ec42adbbd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b121e209131cea2bf59.md>)

[Implementation](<https://github.com/microsoft/LoRA>)

Recorded evaluation configuration: {"Model":"LLaMA2-7b"}. This record does not assert independent reproduction.

### S-MeZO

BoolQ Accuracy: 82.2. Result ID: 699a9efd5698665ec42adbc6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

Recorded evaluation configuration: {"Model":"LLaMA2-7b"}. This record does not assert independent reproduction.

### MeZO-LoRA

BoolQ Accuracy: 80.3. Result ID: 699a9efd5698665ec42adbbf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

Recorded evaluation configuration: {"Model":"LLaMA2-7b"}. This record does not assert independent reproduction.

### AdaZeta

BoolQ Accuracy: 79.8. Result ID: 699a9efd5698665ec42adbc4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75f51e209131cea6aa7f.md>)

Recorded evaluation configuration: {"Model":"LLaMA2-7b"}. This record does not assert independent reproduction.

### ZO-SGD-Adam

BoolQ Accuracy: 78.9. Result ID: 699a9efd5698665ec42adbc2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa734f1e209131cea5bb90.md>)

Recorded evaluation configuration: {"Model":"LLaMA2-7b"}. This record does not assert independent reproduction.

### MeZO

BoolQ Accuracy: 78.8. Result ID: 699a9efd5698665ec42adbbe.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71d41e209131cea5338c.md>)

Recorded evaluation configuration: {"Model":"LLaMA2-7b"}. This record does not assert independent reproduction.

### ZO-AdaMU

BoolQ Accuracy: 78.2. Result ID: 699a9efd5698665ec42adbc3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71ad1e209131cea527d3.md>)

Recorded evaluation configuration: {"Model":"LLaMA2-7b"}. This record does not assert independent reproduction.

### ZO-SGD-Cons

BoolQ Accuracy: 77.1. Result ID: 699a9efd5698665ec42adbc0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa734f1e209131cea5bb90.md>)

Recorded evaluation configuration: {"Model":"LLaMA2-7b"}. This record does not assert independent reproduction.

### R-MeZO

BoolQ Accuracy: 76.7. Result ID: 699a9efd5698665ec42adbc5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

Recorded evaluation configuration: {"Model":"LLaMA2-7b"}. This record does not assert independent reproduction.

### S-MeZO

BoolQ Accuracy: 73.8. Result ID: 699a9efd5698665ec42adbb6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

Recorded evaluation configuration: {"Model":"OPT-13b","Description":"Sparse MeZO"}. This record does not assert independent reproduction.

### R-MeZO

BoolQ Accuracy: 72.3. Result ID: 699a9efd5698665ec42adbb5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

Recorded evaluation configuration: {"Model":"OPT-13b","Description":"MeZO with Random Mask"}. This record does not assert independent reproduction.

### MeZO

BoolQ Accuracy: 72.1. Result ID: 699a9efd5698665ec42adbb4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71d41e209131cea5338c.md>)

Recorded evaluation configuration: {"Model":"OPT-13b"}. This record does not assert independent reproduction.

### ZO-SGD-Sign

BoolQ Accuracy: 68.2. Result ID: 699a9efd5698665ec42adbc1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa734f1e209131cea5bb90.md>)

Recorded evaluation configuration: {"Model":"LLaMA2-7b"}. This record does not assert independent reproduction.

### ICL

BoolQ Accuracy: 66.9. Result ID: 699a9efd5698665ec42adbb3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa680c1e209131cea19d53.md>)

[Implementation](<https://github.com/openai/gpt-3/blob/master/overlap_frequency.md>)

Recorded evaluation configuration: {"Model":"OPT-13b","Full Name":"In-Context Learning"}. This record does not assert independent reproduction.

### Zero Shot

BoolQ Accuracy: 59. Result ID: 699a9efd5698665ec42adbb2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55e46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa680c1e209131cea19d53.md>)

[Implementation](<https://github.com/openai/gpt-3/blob/master/overlap_frequency.md>)

Recorded evaluation configuration: {"Model":"OPT-13b"}. This record does not assert independent reproduction.

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