Contextual_KTO_Mistral_PairRM: benchmark results
Contextual AI Contextual KTO Mistral PairRM is a Mistral-7B-Instruct-v0.2 derivative aligned with three sequential KTO iterations using the Snorkel Mistral PairRM preference dataset. It uses a TuluV2-style prompt format and remains distinct from the Mistral parent and Snorkel DPO-trained releases. Provider: Contextual AI. Access: Open.
Aggregate Rating 1422 ± 36, rank #5086 of 10916 rated models, from 7 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 71.67 | MC2 (%) (0-shot) | 92 |
| Open LLM Leaderboard v1 - HellaSwag | 85.52 | Normalized accuracy (%) (10-shot) | 75 |
| Open LLM Leaderboard v1 - ARC Challenge | 64.76 | Normalized accuracy (%) (25-shot) | 67.9 |
| AlpacaEval 2.0 | 29.71 | LC Win Rate (%) | 67.1 |
| Open LLM Leaderboard v1 - GSM8K | 33.81 | Accuracy (%) (5-shot) | 49.7 |
| Open LLM Leaderboard v1 - MMLU | 60.28 | Accuracy (%) (5-shot) | 47.5 |
| Open LLM Leaderboard v1 - WinoGrande | 75.53 | Accuracy (%) (5-shot) | 42.7 |
Interactive version: theaggregate.ai/model?slug=contextual-kto-mistral-pairrm · How It Works · Data refreshed daily, snapshot 2026-11-08.