The fact that ChatGPT basically turned simple English into a coding language needs to be discussed more.
LLMS
-
Data Selection and Distribution Functions in Large Language Models
By
–
The expectation is replaced by an average over tokens (a few trillion for the largest LMs) so F can be very general. If a human is selecting the y’s among other y’s to create a dataset, then the human is F, hopefully being sensible
-
Learning from Preferences: RLHF, Policy Gradients, and Dagger
By
–
Finally, when learning from preferences, one learns an F(x,y) that enables one to rank and select or do policy gradients (e.g. PPO) as in most RLHF. When the interface allows for corrections (e.g. rewriting the response in a chat agent), then we are in the domain of Dagger.
-
Self-Training: Filtering Functions and Model Ranking Systems
By
–
Self-training: F(x,y) is a filtering/ranking function, eg., what we call a reward/return. The input x may be chosen by humans, but the model generates the y’s and F ranks and selects for further rounds of self-training. F can be explicit or implicit (human in the loop as in RLHF)
-
Learning Methods Unified Through Gradient Optimization Framework
By
–
Funny @sirbayes Learning methods — supervised, RLHF, policy gradients, Dagger, self-training — can be seen as optimisation with the following gradient: grad = Expectation_x,y [ F(x,y) grad log p(y|x) ] Choices of F and how x and y are produced determine the learning type 1/n
-
AI-Powered Long-Form Content and Podcast Summarization
By
–
I want to try it with a realllly long ramble. Would also love to have it listen to podcast convos and write the summary for me.
-
LangChain Tools for Data Loading and AI Chat Applications
By
–
AirByte Loader: https://
langchain.readthedocs.io/en/latest/modu
les/document_loaders/examples/airbyte_json.html
… ChatVectorDBChain Notebook: https://
langchain.readthedocs.io/en/latest/modu
les/chains/combine_docs_examples/chat_vector_db.html
… Chat-Your-Data Challenge: https://
blog.langchain.dev/chat-your-data
-challenge/
… -
Open Source ChatGPT Alternative Dataset Contribution Initiative
By
–
Tenéis toda la info en este vídeo. Pero el TL;DR es que estamos contribuyendo al dataset de un futuro asistente similar a ChatGPT que sea open source. Más datos de calidad Mejor rendimiento. Así que cualquier difusión y ayuda será bienvenida 🙂 https://
youtube.com/watch?v=XYT1Tx
INbuM&ab_channel=DotCSV
… -
Open Assistant surpasses 50000 samples milestone
By
–
ACTUALIZACIÓN DE OPEN ASSISTANT Se ha superado el objetivo de 50.000 muestras propuesto inicialmente. Seguimos! El número de muestras en Español ya supera las 17.000, siendo el 2º mayor idioma del dataset. Empresas, influencers, compañeros… os animáis?
-
QA Over Docs: Diagrams Explaining RAG Architecture
By
–
Some great diagrams here explaining how qa over docs works