claim
For the RACE reading comprehension dataset, models based on Mistral 7B and LLaMA2 produce the most accurate results.
Authors
Sources
- The Hallucinations Leaderboard, an Open Effort to Measure ... huggingface.co via serper
For the RACE reading comprehension dataset, models based on Mistral 7B and LLaMA2 produce the most accurate results.