Unverified60% confidenceFactExact time
Llama 3.3 70B is essentially a fine-tuned version of the earlier Llama 3.1 70B, trained using supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF).
2
Sources
60%
Confidence
Long-term
Relevance
12/8/2024
First Seen
Sources
Llama 3.3 70B In-Depth Review: Testing the Strongest Open-Source LLM with 13 Questions
bilibiliAI-seeker12/8/2024
Related Entities
Related Claims
UnverifiedCodex基于GPT-3的1750亿参数框架，通过监督微调和基于人类反馈的强化学习（RLHF）训练而成70% similarUnverifiedLlama 3提供了8B和70B两个版本，在英文通用任务上的表现已经接近GPT-4级别的闭源模型69% similarUnverifiedStanford's Alpaca project used GPT-3.5 to generate 52K instruction data points for training a 7B-parameter LLaMA.67% similarUnverifiedLLaMA是Meta开源的大语言模型系列，以较小的参数量实现了接近GPT-3.5的性能67% similarUnverifiedLLaMA是Meta开源的大语言模型系列，以较小的参数量实现了接近GPT-3.5的性能67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/665API
curl https://kongchang.com/api/v1/knowledge/claims/665MCP
get_claim(id=665)