WikiText2
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Language Modeling | WikiText2 | Perplexity2.73 | 3,785 | |
| Language Modeling | WikiText2 (val) | Perplexity (PPL)3.03 | 423 | |
| Language Modeling | WikiText2 v1 (test) | Perplexity1.7 | 383 | |
| Language Generation | WikiText2 | Perplexity2.9 | 287 | |
| Language Modeling | Wikitext2 | Perplexity2.58 | 277 | |
| Language Modeling | WikiText2 2016 (test) | Perplexity3.32 | 88 | |
| Language Modeling | WikiText2 (train) | Final Train Loss5.3073 | 16 | |
| Language Modeling | WikiText2 zero-shot | Perplexity26.06 | 13 | |
| Next Token Prediction | Wikitext2 | Perplexity7.64 | 12 | |
| Prompt Reconstruction Defense (TokenInfer attack) | WikiText2 | TRA97.54 | 7 | |
| PII Mitigation and Language Modeling | WikiText2 (test) | Avg PPL531.97 | 3 |
Showing 11 of 11 rows