Jericho Siahaya · Open weights, local
Naluri-279M by Jericho Siahaya scores 50.4 on the System One Index (24th of 24). Calibration error 8.3%.
| Dataset | Accuracy | Rank | Best on the board | Calibration error |
|---|---|---|---|---|
| AG News English | 77.5 | 24th of 24 | 92.5 | 9.9% |
| SST-2 English | 77.0 | 23rd of 24 | 97.0 | 5.3% |
| SST-5 English | 34.0 | 22nd of 24 | 62.5 | 9.1% |
| DAIR Emotion English | 68.5 | 4th of 24 | 91.5 | 20.9% |
| Banking77 English | 31.0 | 20th of 24 | 98.5 | 11.1% |
| BoolQ English | 54.5 | 24th of 24 | 92.5 | 13.6% |
| XNLI English English | 30.7 | 24th of 24 | 93.3 | 40.6% |
| Typed decisions English | 47.3 | 21st of 24 | 79.0 | 7.8% |
| XNLI French French | 31.3 | 24th of 24 | 90.7 | 41.6% |
| XNLI Arabic Arabic | 30.7 | 26th of 26 | 89.3 | 40.9% |
| MASSIVE scenario Arabic | 58.0 | 17th of 26 | 89.5 | 10.1% |
| MASSIVE intent Arabic | 52.0 | 23rd of 26 | 91.5 | 12.6% |
| OSACT4 offensive Arabic | 82.5 | 19th of 26 | 95.5 | 4.1% |
| Tweet sentiment Arabic | 53.0 | 20th of 26 | 80.0 | 6.2% |
| MIRACL rerank Arabic | 28.0 | 26th of 26 | 90.5 | 10.4% |
| MIRACL pairwise Arabic | 50.5 | 26th of 26 | 91.0 | 42.8% |
jerichosiahaya/Naluri-279M-encoderTrained on the train splits of XNLI, MASSIVE and DAIR Emotion; we test their test splits. CC BY-NC 4.0. Run with serve.py from the Naluri repo.
Ranks count the models charted on the leaderboard. Every model gets the same cases and the same request, under the benchmark rules.