# NLU and Question Answering on SuperGLUE Source: [NLU and Question Answering on SuperGLUE]() Collection index: [llms.txt]() Data updated: 2026-03-24T16:08:12.932Z Evaluation of model performance across multiple constituent tasks within the SuperGLUE benchmark suite, covering NLU and Question Answering capabilities, reporting individual task accuracies and an overall average. 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 | NLU and Question Answering | | Dataset (including any stated split) | SuperGLUE | | Metric | SST-2 Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 9 of 9 | | 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 - [SST-2 Accuracy](): Higher is better; unit: PERCENT. - [RTE Accuracy](): Higher is better; unit: PERCENT. - [CB Accuracy](): Higher is better; unit: PERCENT. - [BoolQ Accuracy](): Higher is better; unit: PERCENT. - [WSC Accuracy](): Higher is better; unit: PERCENT. - [WIC Accuracy](): Higher is better; unit: PERCENT. - [SQuAD Accuracy](): Higher is better; unit: PERCENT. - [DROP Accuracy](): Higher is better; unit: PERCENT. - [Average Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | SST-2 Accuracy | Result date | Report paper | | --- | --- | --- | --- | | LoRA | 94.7 | Not recorded | 69c21aa70636e5bf4afc4a8a | | DiZO | 94.4 | Not recorded | 69c21aa70636e5bf4afc4a8a | | CurvZO | 94.1 | Not recorded | 69c21aa70636e5bf4afc4a8a | | FT | 93.8 | Not recorded | 69c21aa70636e5bf4afc4a8a | | CurvZO + LoRA | 93.6 | Not recorded | 69c21aa70636e5bf4afc4a8a | | MeZO + LoRA | 93 | Not recorded | 69c21aa70636e5bf4afc4a8a | | MeZO | 92.1 | Not recorded | 69c21aa70636e5bf4afc4a8a | | DiZO + LoRA | 92 | Not recorded | 69c21aa70636e5bf4afc4a8a | | Zero-shot | 56.3 | Not recorded | 69c21aa70636e5bf4afc4a8a | ### LoRA SST-2 Accuracy: 94.7. Result ID: 69c2b6d947f34ce924ac0a2a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model":"OPT-2.7B","Number of training samples":"1,000","Evaluation protocol":"LoRA"}. This record does not assert independent reproduction. ### DiZO SST-2 Accuracy: 94.4. Result ID: 69c2b6d947f34ce924ac0a2c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"OPT-2.7B","Number of training samples":"1,000","Evaluation protocol":"Zeroth-Order (ZO)"}. This record does not assert independent reproduction. ### CurvZO SST-2 Accuracy: 94.1. Result ID: 69c2b6d947f34ce924ac0a2d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"OPT-2.7B","Number of training samples":"1,000","Evaluation protocol":"Zeroth-Order (ZO)"}. This record does not assert independent reproduction. ### FT SST-2 Accuracy: 93.8. Result ID: 69c2b6d947f34ce924ac0a29. [Reporting paper]() Recorded evaluation configuration: {"Model":"OPT-2.7B","Number of training samples":"1,000","Evaluation protocol":"Full Fine-Tuning"}. This record does not assert independent reproduction. ### CurvZO + LoRA SST-2 Accuracy: 93.6. Result ID: 69c2b6d947f34ce924ac0a30. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"OPT-2.7B","Number of training samples":"1,000","Evaluation protocol":"ZO + PEFT"}. This record does not assert independent reproduction. ### MeZO + LoRA SST-2 Accuracy: 93. Result ID: 69c2b6d947f34ce924ac0a2e. [Reporting paper]() Recorded evaluation configuration: {"Model":"OPT-2.7B","Number of training samples":"1,000","Evaluation protocol":"ZO + PEFT"}. This record does not assert independent reproduction. ### MeZO SST-2 Accuracy: 92.1. Result ID: 69c2b6d947f34ce924ac0a2b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"OPT-2.7B","Number of training samples":"1,000","Evaluation protocol":"Zeroth-Order (ZO)"}. This record does not assert independent reproduction. ### DiZO + LoRA SST-2 Accuracy: 92. Result ID: 69c2b6d947f34ce924ac0a2f. [Reporting paper]() Recorded evaluation configuration: {"Model":"OPT-2.7B","Number of training samples":"1,000","Evaluation protocol":"ZO + PEFT"}. This record does not assert independent reproduction. ### Zero-shot SST-2 Accuracy: 56.3. Result ID: 69c2b6d947f34ce924ac0a28. [Reporting paper]() Recorded evaluation configuration: {"Model":"OPT-2.7B","Number of training samples":"1,000","Evaluation protocol":"Zero-shot"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.