1、LLM的规模化定律 – 大就是好。大语言模型在下一个单词预测任务的准确性方面的表现是一个非常平滑且可预测的仅两个变量的函数。只需要知道的神经网络中的参数量(n)和训练的文本量(d),就可以非常有信心地预估LLM的性能,而且这条曲线似乎并没有见顶的迹象。也就是说,当前不必考虑算法的优化,单纯靠加大参数...
2、当模型开始使用工具 - 类似人类能通过大量使用工具来解决各种各样的任务一样,给与LLMs各种工具能极大加强其实用性,例如上网浏览、计算器、编程和调用其他模型,最新的ChatGPT+Dalle3就是个很好的例子(下图是ChatGPT绘制的时间穿梭 - 星巴克的变迁之路,从不存在星巴克的1920年到刚开店的1970,到大家熟悉的2020年,再...
https://medium.com/@yianyao1994/llm-alignments-part-1-overview-05515efa3bb1 对齐方法的大概介绍,主要点是梳理了下面这张图: part2:RLHF https://medium.com/@yianyao1994/llm-alignments-part-2-rlhf-and-rlaif-0dba2b5a9423 第一步:1.3W 条数据,就可以搞SFT。(质量比数量更重要。) 第二步:关键...
Unlearning Sensitive Content from LLMs (9 minute read) Mar 31|AI xAI acquires X in $80B all-stock deal (1 minute read) Mar 31|AI Gemini 2.5: Our most intelligent AI model (1 minute read) Mar 31|AI Reasoning augmented generation code (GitHub Repo) ...
Youtube for people with no patience-aka TLDR youtube with help of LLM to speed up the information gathering and learnining - WenchenLi/TLDR-youtube-LLM
reinforcement-learning tldr supervised-learning fine-tuning document-summarization llm retrieval-augmented-generation llama3 Updated Feb 2, 2025 Python giocoal / reddit-tldr-summarizer-and-topic-modeling Star 7 Code Issues Pull requests Discussions Extreme Extractive Text Summarization and Topic Modelin...
o1 通过增强 LLM 文本模型的逻辑推理能力,解锁更多复杂应用,进而提升整体大模型的认知水平。 ③ 如果 o1 模型的能力不断提升,可以反哺 GPT-4o,通过替换基座模型、生成合成数据或蒸馏模型等方式,提升 GPT-4o 的复杂任务解决能力。 2、对于业界有关 o1 对于 Scaling Law 的影响的讨论,张俊林和北大对齐团队也在...
we’re sharing details on two versions of our24,576-GPU data center scale cluster at Meta. These clusters support our current and next generation AI models, including Llama 3, the successor toLlama 2, our publicly released LLM, as well as AI research and development across GenAI and...
Docker Desktop 4.40: Model Runner to run LLMs locally, more powerful Docker AI Agent, and expanded AI Tools Catalog By Yiwen Xu April 1, 2025 8 Ways to Empower Engineering Teams to Balance Productivity, Security, and Innovation By Lance Haig March 25, 2025 Leveraging Docker with TensorFlow Mo...
they want to build stuff on top of LLMs and you talk about all the startups that’s sprouted up in the post-ChatGPT moment. Is the explosion of innovation — is this going to be in a fundamentally different layer of the stack going forward, just further up on top and t...