Tülu 3
Tülu 3 是艾伦人工智能研究所（AI2）开源的后训练方法与模型系列，公开了完整的 SFT、DPO、RLVR 训练配方（open-instruct），让社区可以复现 SOTA 级后训练效果。Tülu 3 系列在各参数量级别均表现领先，是开源大模型研究的标杆项目。 🔗 GitHub: https://github.com/allenai/open-instruct | GitHub Stars: 3.8k+ | 🔗 模型: https://huggingface.co/allenai/Llama-3.1-Tulu-3-8B
Tülu 3 from AI2 is an open post-training recipe and model family. It publishes the complete SFT, DPO, and RLVR training pipeline (open-instruct) so the community can reproduce frontier-style results, achieving state-of-the-art performance at every model size. 🔗 GitHub: https://github.com/allenai/open-instruct | GitHub Stars: 3.8k+ | 🔗 Model: https://huggingface.co/allenai/Llama-3.1-Tulu-3-8B
📋 Product Info
🎯 Why This Tool
Tülu 3 belongs to the Open Source LLMs category on hedirbase. Tagged with: open-source-llm,post-training,allenai,research.
Tülu 3 from AI2 is an open post-training recipe and model family. It publishes the complete SFT, DPO, and RLVR training pipeline (open-instruct) so the community can reproduce frontier-style results,