DCLM
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Language Modeling | DCLM (val) | Loss2.589 | 78 | |
| Zero-shot Evaluation | DCLM CORE V2 | CORE_V2 Score48 | 17 | |
| Language Model Evaluation | DCLM Core | DCLM Core Score49.3 | 12 | |
| Language Modeling | DCLM | Loss1.863 | 11 | |
| Zero-shot Evaluation | DCLM CORE | CORE Score0.268 | 9 | |
| Natural Language Understanding | DCLM | DCLM Core Score48.9 | 9 | |
| Language Modeling | DCLM benchmark | Macro Avg Value14.58 | 9 | |
| General Language Capability | DCLM CORE v2 (test) | Commonsense Score61.9 | 7 | |
| Language Understanding | DCLM evaluation suite (test) | HellaSwag Accuracy62.9 | 7 | |
| Aggregate Evaluation | DCLM CORE (val) | CORE27.2 | 5 | |
| Scaling Law Fitting | DCLM strongly regularized baseline 20 points (Full fit) | RMSE0.008 | 4 | |
| Zero-shot Language Evaluation | DCLM Pro | WinoGrande57.93 | 2 |
Showing 12 of 12 rows