We are making big strides towards programmatic alignment! Using a reward model w/ DPO, Snorkel AI ranked 2nd on the AlpacaEval 2.0 LLM leaderboard using only a 7B model! Hear from the researcher behind this achievement at our Enterprise LLM Summit tomorrow! #SnorkelAI #AI #LLMs
LLMS
-
New features for managing personal data in ChatGPT and Custom GPTs
By
–
It will be possible to see what GPT learned about you and remove this information. There is also a dropdown which only has a ChatGPT option as of now but it looks like you will be able to browse and manage the same for Custom GPTs
-
ChatGPT Developing New Personalisation Memory Management Options
By
–
ChatGPT is working on an option to manage Personalisation memory
-
Key Research Topics for Improving RAG Accuracy in LLM Applications
By
–
A good custom LLM application typically involves RAG and while naive RAG is easy to implement, it rarely performs well. Here are key areas research topics in RAG that are essential for increasing accuracy of custom LLM apps In-Context Learning – LLMs do poorly when you send a
-
LangChain QA with RAG and Advanced Retrieval Techniques
By
–
two pages id check out: QA with RAG, a page covering lot of different concept related to this use case: https://
python.langchain.com/docs/use_cases
/question_answering/
… Retrieval: a page covering advanced retrieval techniques: https://
python.langchain.com/docs/modules/d
ata_connection/retrievers/
… -
Kit Kat’s AI Break Study: Marketing or Misinformation?
By
–
Kit Kat Canada just used a @GoogleDeepMind study to advertise chocolate.
— Allie K. Miller (@alliekmiller) 24 janvier 2024
Because if AI performs better after a break, maybe you need one too 🍫
(Friendly reminder that LLMs are not truth tellers and reducing hallucinations does not mean you should trust the output blindly.) pic.twitter.com/HufEma2UmDKit Kat Canada just used a @GoogleDeepMind study to advertise chocolate. Because if AI performs better after a break, maybe you need one too (Friendly reminder that LLMs are not truth tellers and reducing hallucinations does not mean you should trust the output blindly.)
-
Automated Testing for LLMOps: CI/CD Best Practices
By
–
New short course on Automated Testing for LLMOps, by @CircleCI's CTO Rob Zuber! This teaches you how to adapt some key ideas from CI (continuous integration), which has been a pillar of efficient software engineering, to building LLM-based applications.
— Andrew Ng (@AndrewYNg) 24 janvier 2024
Tweaking an LLM-based app… pic.twitter.com/95kQkbAZsGNew short course on Automated Testing for LLMOps, by @CircleCI
's CTO Rob Zuber! This teaches you how to adapt some key ideas from CI (continuous integration), which has been a pillar of efficient software engineering, to building LLM-based applications. Tweaking an LLM-based app -
Fine-Tuning Foundation Models: New Safety Risks Emerge
By
–
When users of foundation models customize them through a process called fine-tuning, new safety risks emerge. Read the HAI Policy Brief:
-
Fireworks AI Partnership Brings Free Smaller Models to Playground
By
–
We will look into it! in the meantime, we've partnered with @FireworksAI_HQ to bring a lot of those smaller models to the playground for free
-
Groq Congratulates OpenAI on ChatGPT Store Launch
By
–
Hey @SamA
, Congrats on the #ChatGPT store! We had fun poking around, alas is was slow poking. We recommend supercharging it with 300T/s of #GroqSpeed! https://
groq.com/hey-sam/ #Groq ® #LLM #GenAI