pancho-v1-qw25-3B-UNAMGS: benchmark results
fblgit Qwen 2.5 3B Instruct fine-tune trained on Magpie data with MGS and UNA interventions at selected MLP layers. Provider: fblgit. Access: Open.
Aggregate Rating 1461 ± 31, rank #3262 of 10916 rated models, from 56 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Japanese LLM - Mbpp Pylint Check | 80.92 | Score (%) | 69 |
| Open Japanese LLM - CG | 41.57 | Score (%) | 65.1 |
| Open LLM Leaderboard - IFEval | 53.61 | Score | 64 |
| Open LLM Leaderboard - MMLU-Pro | 37.66 | Score | 63.3 |
| Open LLM Leaderboard - MATH Level 5 | 15.71 | Score | 63.2 |
| Open Japanese LLM - Wiki Coreference SET F1 | 3.65 | Score (%) | 58.7 |
| Open LLM Leaderboard - GPQA | 29.7 | Score | 52.5 |
| Open Japanese LLM - Jnli Exact Match | 79.7 | Score (%) | 50.7 |
| Open LLM Leaderboard - BBH | 49.26 | Score | 45.4 |
| Open LLM Leaderboard - MuSR | 40.27 | Score | 44.3 |
| Open Japanese LLM - Wiki NER SET F1 | 3.54 | Score (%) | 39.4 |
| Open Japanese LLM - Jsick Exact Match | 72.25 | Score (%) | 36.9 |
Interactive version: theaggregate.ai/model?slug=pancho-v1-qw25-3b-unamgs · How It Works · Data refreshed daily, snapshot 2026-11-08.