Release v1.4
🚀Major Update: Introducing WizardLM-70B-V1.0 trained from Llama-2
Compared with Llama-2-70b-chat, there are the following updates:
- AlpacaEval Leaderboard: 92.66% -> 92.91%
- MT-Bench Leaderboard: 6.86 -> 7.78
- Gsm8K: 56.8% -> 77.6%
- HumanEval: 32.3 -> 50.6