Llama-3.1-Tulu-3-8B-RM: benchmark results
Provider: Allen AI. Released 2024-11-21. Access: Open.
Aggregate Rating 1495 ± 31, rank #801 of 1656 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| RewardBench 2 Factuality | 74.53 | Accuracy (%) | 80.4 |
| RubricBench - Code | 53.5 | Preference accuracy (%) on the code pairs on RubricBench's 1 | 77.8 |
| RubricBench - Chat | 45.1 | Preference accuracy (%) on the chat pairs on RubricBench's 1 | 66.7 |
| RewardBench 2 Math | 64.48 | Score (%) | 63.5 |
| RubricBench - Instruction Following | 41.1 | Preference accuracy (%) on the instruction-following pairs o | 50 |
| RewardBench 2 Ties | 52.43 | Score (%) | 45.9 |
| RMGAP | 62.68 | Pairwise accuracy (%; chosen response scored above a rejecte | 37.5 |
| RMGAP - Best-of-4 | 39.26 | Best-of-4 accuracy (%; chosen response scored above all thre | 37.5 |
| RubricBench | 47.1 | Preference accuracy (%) on RubricBench's 1,147 preference pa | 35.7 |
| RewardBench 2 | 59 | Score (%) | 33.7 |
| RubricBench - Safety | 43.8 | Preference accuracy (%) on the safety pairs on RubricBench's | 33.3 |
| RewardBench 2 Precise IF | 34.69 | Score (%) | 29 |
Interactive version: theaggregate.ai/model?slug=llama-3-1-tulu-3-8b-rm · How It Works · Data refreshed daily, snapshot 2026-10-20.