MoE_13B_DPO: benchmark results
yunconglong's DPO fine-tune of their Truthful DPO FusionNet 7Bx2 MoE 13B model, trained on the Intel Orca DPO pairs dataset. Provider: yunconglong. Released 2024-01-28. Access: Open.
Aggregate Rating 1477 ± 24, rank #854 of 1635 rated models, from 19 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - WinoGrande | 88 | Accuracy (%) (5-shot) | 99.8 |
| Open LLM Leaderboard v1 - ARC Challenge | 74.32 | Normalized accuracy (%) (25-shot) | 99.2 |
| Open LLM Leaderboard v1 - HellaSwag | 89.39 | Normalized accuracy (%) (10-shot) | 99.2 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 78.47 | MC2 (%) (0-shot) | 99.1 |
| Open Chinese LLM - TruthfulQA MC | 62.61 | Accuracy (%) | 91.4 |
| Open LLM Leaderboard v1 - GSM8K | 67.63 | Accuracy (%) (5-shot) | 84 |
| Open Chinese LLM - WinoGrande | 67.48 | Accuracy (%) | 83.4 |
| Open Portuguese LLM - ASSIN2 RTE | 92.57 | Macro F1 (%) | 80.4 |
| Open Portuguese LLM - FaQuAD NLI | 77.36 | Macro F1 (%) | 79.5 |
| Open LLM Leaderboard v1 - MMLU | 64.48 | Accuracy (%) (5-shot) | 71.1 |
| Open Chinese LLM - HellaSwag | 61.77 | Accuracy (%) | 64.8 |
| Open Chinese LLM - ARC Challenge | 54.1 | Accuracy (%) | 63.3 |
Interactive version: theaggregate.ai/model?slug=moe-13b-dpo · How It Works · Data refreshed daily, snapshot 2026-11-05.