Clown-DPO-Extended: benchmark results
A 40-layer block expansion by Arcee AI of CorticalStack/pastiche-crown-clown-7b-dare-dpo. The merge recipe inserts blocks with zero-scaled output and down projections; the release card says the expanded checkpoint has not undergone fine-tuning. Provider: Arcee AI. Access: Open.
Aggregate Rating 1466 ± 43, rank #2900 of 10870 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 78.78 | MC2 (%) (0-shot) | 99.4 |
| Open LLM Leaderboard v1 - HellaSwag | 89.09 | Normalized accuracy (%) (10-shot) | 96.9 |
| Open LLM Leaderboard v1 - ARC Challenge | 73.12 | Normalized accuracy (%) (25-shot) | 95.8 |
| Open LLM Leaderboard v1 - WinoGrande | 84.69 | Accuracy (%) (5-shot) | 94.4 |
| Open LLM Leaderboard v1 - GSM8K | 68.23 | Accuracy (%) (5-shot) | 85.5 |
| Open LLM Leaderboard v1 - MMLU | 64.52 | Accuracy (%) (5-shot) | 72 |
Interactive version: theaggregate.ai/model?slug=clown-dpo-extended · How It Works · Data refreshed daily, snapshot 2026-11-12.