# Language Modeling on WikiText (WT) Source: [Language Modeling on WikiText (WT)]() Collection index: [llms.txt]() Data updated: 2026-04-07T10:38:41.416Z Evaluation of Language Modeling performance on the WikiText (WT) dataset, specifically measuring the percentage change in Perplexity (PPL) relative to a pretrained baseline model. 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 | Language Modeling | | Dataset (including any stated split) | WikiText (WT) | | Metric | Relative PPL Change (%) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 16 of 16 | | 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 - [Relative PPL Change (%)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Relative PPL Change (%) | Result date | Report paper | | --- | --- | --- | --- | | LoRA | 31 | Not recorded | 69d48fb47d0172140ec744ce | | LoRA | 25.2 | Not recorded | 69d48fb47d0172140ec744ce | | LoRA | 23.6 | Not recorded | 69d48fb47d0172140ec744ce | | LoRA | 14.5 | Not recorded | 69d48fb47d0172140ec744ce | | GAIN | 11.1 | Not recorded | 69d48fb47d0172140ec744ce | | GAIN | 11 | Not recorded | 69d48fb47d0172140ec744ce | | GAIN | 11 | Not recorded | 69d48fb47d0172140ec744ce | | GAIN | 10.6 | Not recorded | 69d48fb47d0172140ec744ce | | GAIN | 10.6 | Not recorded | 69d48fb47d0172140ec744ce | | GAIN | 10.3 | Not recorded | 69d48fb47d0172140ec744ce | | GAIN | 10 | Not recorded | 69d48fb47d0172140ec744ce | | LoRA | 9.8 | Not recorded | 69d48fb47d0172140ec744ce | | LoRA | 8.4 | Not recorded | 69d48fb47d0172140ec744ce | | LoRA | 1.2 | Not recorded | 69d48fb47d0172140ec744ce | | LoRA | 1 | Not recorded | 69d48fb47d0172140ec744ce | | GAIN | 0 | Not recorded | 69d48fb47d0172140ec744ce | ### LoRA Relative PPL Change (%): 31. Result ID: 69d4deac47f34ce924bd611a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+HS"}. This record does not assert independent reproduction. ### LoRA Relative PPL Change (%): 25.2. Result ID: 69d4deac47f34ce924bd6114. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+WT"}. This record does not assert independent reproduction. ### LoRA Relative PPL Change (%): 23.6. Result ID: 69d4deac47f34ce924bd6119. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+SST"}. This record does not assert independent reproduction. ### LoRA Relative PPL Change (%): 14.5. Result ID: 69d4deac47f34ce924bd6118. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+LB"}. This record does not assert independent reproduction. ### GAIN Relative PPL Change (%): 11.1. Result ID: 69d4deac47f34ce924bd610d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+Fin"}. This record does not assert independent reproduction. ### GAIN Relative PPL Change (%): 11. Result ID: 69d4deac47f34ce924bd610c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+WT"}. This record does not assert independent reproduction. ### GAIN Relative PPL Change (%): 11. Result ID: 69d4deac47f34ce924bd610e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+PG"}. This record does not assert independent reproduction. ### GAIN Relative PPL Change (%): 10.6. Result ID: 69d4deac47f34ce924bd610f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+MdQ"}. This record does not assert independent reproduction. ### GAIN Relative PPL Change (%): 10.6. Result ID: 69d4deac47f34ce924bd6111. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+SST"}. This record does not assert independent reproduction. ### GAIN Relative PPL Change (%): 10.3. Result ID: 69d4deac47f34ce924bd6112. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+HS"}. This record does not assert independent reproduction. ### GAIN Relative PPL Change (%): 10. Result ID: 69d4deac47f34ce924bd6110. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+LB"}. This record does not assert independent reproduction. ### LoRA Relative PPL Change (%): 9.8. Result ID: 69d4deac47f34ce924bd6116. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+PG"}. This record does not assert independent reproduction. ### LoRA Relative PPL Change (%): 8.4. Result ID: 69d4deac47f34ce924bd6115. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+Fin"}. This record does not assert independent reproduction. ### LoRA Relative PPL Change (%): 1.2. Result ID: 69d4deac47f34ce924bd6117. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+MdQ"}. This record does not assert independent reproduction. ### LoRA Relative PPL Change (%): 1. Result ID: 69d4deac47f34ce924bd6113. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+Med"}. This record does not assert independent reproduction. ### GAIN Relative PPL Change (%): 0. Result ID: 69d4deac47f34ce924bd610b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"GPT-2 Large (774M)","Learning rate":"10^-3","Training sequence stage":"+Med"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.