Happy New Year, and welcome to Athina's weekly newsletter llmEngineer.weekly – a curated digest for LLM Engineers. In this newsletter, we'll share industry news, interesting readings, useful libraries, and some pro tips for LLM development every week.
TLDR: What's in this week's issue of llmEngineer.weekly
- Industry Reads: AI 2023 Timeline, AI x Robotics, Apple’s new LLM
- Paper: 32 methods for LLM hallucination mitigation
- Useful Library: Open Source Library for Swappable LLMs
- A Pro(mpting) Tip to get structured output from an LLM
Reading List
Industry Developments
- The Mixtral 8x7B model is supposed to be the best open-source model on the market, with performance comparable to GPT-3.5-turbo
- Timeline of AI in 2023
- AI x Robotics
- OpenAI models being deprecated this week
- Apple’s Ferret LLM + Hosting LLMs on-device
Paper of the Week
32 methods for LLM Hallucination Mitigation
A recent paper has analyzed 32 different methods to reduce hallucinations in LLM outputs. These methods range from different prompt engineering techniques to prompt tuning and fine tuning.