NeuralSynthesis-7B-v0.1: benchmark results
Kukedlc's model-stock merge of four Mistral-7B variants on a YamshadowExperiment28 base, built with LazyMergekit. Provider: Kukedlc. Released 2024-04-06. Access: Open.
Aggregate Rating 1454 ± 16, rank #974 of 1656 rated models, from 35 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 78.15 | MC2 (%) (0-shot) | 98.6 |
| Open LLM Leaderboard v1 - HellaSwag | 89.18 | Normalized accuracy (%) (10-shot) | 98.3 |
| Open Korean LLM - KorNAT-CKA | 38.9 | Normalized accuracy (%) | 98.2 |
| Open Korean LLM - KorNAT-Harmless | 76.74 | Normalized accuracy (%) | 97.8 |
| Open LLM Leaderboard v1 - WinoGrande | 85.24 | Accuracy (%) (5-shot) | 97.8 |
| Open LLM Leaderboard v1 - ARC Challenge | 73.04 | Normalized accuracy (%) (25-shot) | 95.2 |
| Open Chinese LLM - TruthfulQA MC | 63.77 | Accuracy (%) | 94.7 |
| Open LLM Leaderboard v1 - GSM8K | 70.81 | Accuracy (%) (5-shot) | 94.4 |
| Open Korean LLM Leaderboard | 41.41 | Average Score (%) | 82.7 |
| Open Ko-LLM Leaderboard | 41.41 | Average (%) | 82.6 |
| Open Portuguese LLM - FaQuAD NLI | 78.07 | Macro F1 (%) | 82.6 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 43.82 | Exact match, flexible extract (%) | 78.7 |
Interactive version: theaggregate.ai/model?slug=neuralsynthesis-7b-v0-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.