AI knowledge Page 10

  Transformer is a deep learning modeling architecture for Natural Language Processing (NLP), proposed by Vaswani et al. in 2017. It is mainly used for processing sequence-to-sequence tasks, such as machine translation, text generation, and so on. Briefly, the original Transformer model for text generation...

DISCLAIMER: While basic hinting techniques (e.g., zero/few sample examples or imperative hints) are very efficient, more sophisticated hints may be more effective when faced with some complex puzzles (e.g., math/programming or problems requiring multi-step logical reasoning). Since Large Language Models (LLMs) deal with such problems...

Title: Agent AI Towards a Holistic Intelligence Author: Qiuyuan Huang, Naoki Wake, Bidipta Sarkar, Zane Durante, Ran Gong, Rohan Taori, Yusuke Noda, Demetri Terzopoulos, Noboru Kuno, Ade Famoti, Ashle...

Paper address: https://arxiv.org/abs/2404.17723 Knowledge Graphs can only extract entity relations in a targeted manner, and such stably extractable entity relations can be understood as close to structured data. Figure 1 illustrates a combination of Knowledge Graph (KG) and Retrieval Augmented Generation (RAG)...

The following focuses on the basic idea of hint engineering and how it can improve the performance of the Large Language Model (LLM)... Interfaces for LLM: One of the key reasons why large language models are so hot is that their text-to-text interfaces enable a minimalist operational experience. In the past, solving tasks using deep learning typically required...

Open source address: https://github.com/cpacker/MemGPT Thesis address: https://arxiv.org/abs/2310.08560 Official website: https://memgpt.ai/ MemGPT supports: 1. Management of long-term memory or state 2. Linking of RAG-based technologies External data sources 3.

Researchers have investigated a "jailbreak attack" technique - a method that can be used to bypass security fences set up by developers of large language models (LLMs). The technique, known as the "multisample jailbreak attack," works on Anthropic's own models as well as those produced by other AI companies. The researchers pre...

Original: https://arxiv.org/pdf/2210.03629.pdf Can't understand how ReAct works and applies even after reading it? See ReAct Implementation Logic in Practice with real-world examples. Abstract Although large-scale language models (LLMs) are useful in the tasks of language understanding and interactive decision...

RAG (Retrieve Augmented Generation) is a technique for optimizing the output of Large Language Models (LLMs) based on authoritative knowledge base information. This technique extends the functionality of LLMs to refer to the internal knowledge base of a particular domain or organization when generating responses to...

Original text: "Dense X Retrieval: What Retrieval Granularity Should We Use?" Note: This method is suitable for a small number of models, such as the OPENAI series, the Claude series, Mixtral, Yi, and qwen. Abstract In open-domain natural language processing (NLP) tasks, ...

Today I read an interesting paper "Large Language Models as Analogical Reasoners", which mentions a new approach to Prompt - "Analogical Prompting. If you are familiar with prompt engineering, you must have heard of "Chain of Thought" (CoT), which is an analogical method of prompting...

Abstract The reasoning performance of Large Language Models (LLMs) on a wide range of problems relies heavily on chained-thinking prompts, which involves providing a number of chained-thinking demonstrations as exemplars in the prompts. Recent research, e.g., thinking trees, has pointed to exploration and self-assessment of reasoning in complex problem solving ...

TL;DR This 8200+ word article takes about 15 minutes to read in full. This article briefly reviews the history of ChatGPT, the latest big-model application of deep learning, from perceptual machines to ... Original article: https://hutusi.com/articles/the-history-of-neural-networks There is nothing to fear in life, only...