Report center

计算机行业：大语言模型的前世、今生与未来-20230831-财通证券

大语言模型：NLP技术的奇点时刻。语言模型的核心是对任意一段文本序列进行概率建模，用一个高维向量来表示一个token 的全部特征。我们将深度学习应用于NLP领域的范式转移总结为"有监督的机器学习->预训练+微调->预训练+提示"。其中，ChatGPT的横空出世开启了“预训练+提示"的新范式，这主要得益于OpenAI对生成类模型（GPT）和算法规模化（Scalability）这两个基础技术路线的成功

No subject companies

Publishers

财通证券 View content →

Suggested next hops

Ranked by edge type and evidence strength (max 5). Prefer these over scrolling the full lists.

1财通证券Published by or linked to this institution1
2LLMs & AIGCSame industry · 1 related reports1
3What Does Venture Capital Actually Finance? Scale and Structure around the WorldSame industry · 1 related reports1
42025 State of AI Report: The Builder’s PlaybookSame industry · 1 related reports1
5How to Build a Reader: Evidence from a Scalable Literacy Intervention in GhanaSame industry · 1 related reports1

Ask about this report neighborhood Read full report →

Other hot industries

LLMs & AIGC View content →

Similar reports

What Does Venture Capital Actually Finance? Scale and Structure around the World View content →
2025 State of AI Report: The Builder’s Playbook View content →
How to Build a Reader: Evidence from a Scalable Literacy Intervention in Ghana View content →
When Digital Nudges Nag: Evidence on Program Take-up from a WhatsApp Intervention View content →
Technology Facilitated Gender Based Violence in the Middle East, North Africa, Afghanistan & Pakistan: Logging Women ou…View content →
Competing in the Face of Climate Risks: Evidence from Firms and Policy Priorities in MENAAP View content →

← Back to knowledge graph