# Natural Language Understanding on GLUE (test) (MNLI, QQP, SST-2, SST-5) Source: [Natural Language Understanding on GLUE (test) (MNLI, QQP, SST-2, SST-5)]() Collection index: [llms.txt]() Data updated: 2026-05-19T10:12:11.631Z Evaluation of general language understanding capabilities on the GLUE test set, reporting scores specifically for the MNLI, QQP, SST-2, and SST-5 tasks. 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 (test) | | Metric | MNLI Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 30 of 30 | | 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 - [MNLI Score](): Higher is better; unit: PERCENT. - [QQP Score](): Higher is better; unit: PERCENT. - [SST-2 Accuracy](): Higher is better; unit: PERCENT. - [SST-5 Accuracy](): Higher is better; unit: PERCENT. - [Average GLUE Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | MNLI Score | Result date | Report paper | | --- | --- | --- | --- | | DP-SelFT+LoRA | 82.2 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | LoRA | 81.2 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT+LoRA | 80.5 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT | 80.3 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | LoRA | 79.8 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT | 79.7 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Full Parameter | 78.2 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Heuristic-based | 77.9 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Full Parameter | 77.7 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Heuristic-based | 77.6 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT | 77 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Full Parameter | 76.3 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Heuristic-based | 76.1 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | LoRA | 74.6 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT+LoRA | 74.5 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT | 59 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Full Parameter | 55.5 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT | 54.4 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Full Parameter | 47.5 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Heuristic-based | 46.5 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Heuristic-based | 42.7 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT+LoRA | 41.4 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Full Parameter | 40.8 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT | 40.2 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT+LoRA | 38.8 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | LoRA | 38.7 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | Heuristic-based | 36.8 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | LoRA | 36.1 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | DP-SelFT+LoRA | 35.6 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | | LoRA | 34.5 | Not recorded | 6a0bee9fdbc836b87d82c4b9 | ### DP-SelFT+LoRA MNLI Score: 82.2. Result ID: 6a0c3734fbc373592c02de66. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"10","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### LoRA MNLI Score: 81.2. Result ID: 6a0c3734fbc373592c02de63. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"10","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### DP-SelFT+LoRA MNLI Score: 80.5. Result ID: 6a0c3734fbc373592c02de5c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"5","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### DP-SelFT MNLI Score: 80.3. Result ID: 6a0c3734fbc373592c02de65. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"10","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### LoRA MNLI Score: 79.8. Result ID: 6a0c3734fbc373592c02de59. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"5","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### DP-SelFT MNLI Score: 79.7. Result ID: 6a0c3734fbc373592c02de5b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"5","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### Full Parameter MNLI Score: 78.2. Result ID: 6a0c3734fbc373592c02de62. [Reporting paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"10","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### Heuristic-based MNLI Score: 77.9. Result ID: 6a0c3734fbc373592c02de64. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"10","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### Full Parameter MNLI Score: 77.7. Result ID: 6a0c3734fbc373592c02de58. [Reporting paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"5","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### Heuristic-based MNLI Score: 77.6. Result ID: 6a0c3734fbc373592c02de5a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"5","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### DP-SelFT MNLI Score: 77. Result ID: 6a0c3734fbc373592c02de51. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"1","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### Full Parameter MNLI Score: 76.3. Result ID: 6a0c3734fbc373592c02de4e. [Reporting paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"1","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### Heuristic-based MNLI Score: 76.1. Result ID: 6a0c3734fbc373592c02de50. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"1","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### LoRA MNLI Score: 74.6. Result ID: 6a0c3734fbc373592c02de4f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"1","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### DP-SelFT+LoRA MNLI Score: 74.5. Result ID: 6a0c3734fbc373592c02de52. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"1","Backbone":"RoBERTa-Large"}. This record does not assert independent reproduction. ### DP-SelFT MNLI Score: 59. Result ID: 6a0c3734fbc373592c02de6a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"10","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### Full Parameter MNLI Score: 55.5. Result ID: 6a0c3734fbc373592c02de67. [Reporting paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"10","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### DP-SelFT MNLI Score: 54.4. Result ID: 6a0c3734fbc373592c02de60. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"5","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### Full Parameter MNLI Score: 47.5. Result ID: 6a0c3734fbc373592c02de5d. [Reporting paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"5","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### Heuristic-based MNLI Score: 46.5. Result ID: 6a0c3734fbc373592c02de69. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"10","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### Heuristic-based MNLI Score: 42.7. Result ID: 6a0c3734fbc373592c02de5f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"5","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### DP-SelFT+LoRA MNLI Score: 41.4. Result ID: 6a0c3734fbc373592c02de6b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"10","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### Full Parameter MNLI Score: 40.8. Result ID: 6a0c3734fbc373592c02de53. [Reporting paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"1","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### DP-SelFT MNLI Score: 40.2. Result ID: 6a0c3734fbc373592c02de56. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"1","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### DP-SelFT+LoRA MNLI Score: 38.8. Result ID: 6a0c3734fbc373592c02de61. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"5","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### LoRA MNLI Score: 38.7. Result ID: 6a0c3734fbc373592c02de68. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"10","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### Heuristic-based MNLI Score: 36.8. Result ID: 6a0c3734fbc373592c02de55. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"1","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### LoRA MNLI Score: 36.1. Result ID: 6a0c3734fbc373592c02de5e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"5","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### DP-SelFT+LoRA MNLI Score: 35.6. Result ID: 6a0c3734fbc373592c02de57. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"1","Backbone":"OPT-350M"}. This record does not assert independent reproduction. ### LoRA MNLI Score: 34.5. Result ID: 6a0c3734fbc373592c02de54. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Privacy Budget (epsilon)":"1","Backbone":"OPT-350M"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.