Open-domain Conversation on NO-ConvAI2 NLEBench (test)
4.28BLEUNorGPT-23B
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| NorGPT-23BParameters=23B2026.01 | 4.28 | 16.72 | 15.95 | 84.41 | 0.64 | |
| NorGPT-3BParameters=3B2026.01 | 4.14 | 17.09 | 16.33 | 84.68 | 0.87 | |
| NB-GPT-J-6BParameters=6B, Backbone=GPT-J2026.01 | 3.87 | 17.05 | 16.26 | 85.83 | 0.68 | |
| NorLlama-3BParameters=3B, Backbone=Llama2026.01 | 3.82 | 15.2 | 14.53 | 82.47 | 0.61 | |
| NorGPT-3B-continueParameters=3B, Training=Continued Pre-training2026.01 | 3.63 | 16.47 | 15.73 | 86.33 | 0.71 | |
| NorGPT-369MParameters=369M2026.01 | 3.37 | 16.94 | 16.21 | 86.54 | 0.56 | |
| GPT-3.5Model=GPT-3.5-Turbo2026.01 | 2.14 | 10.82 | 9.96 | 85.8 | 0.72 |