Small update: open-weight models are SOTA for the first time in history
OPEN SOURCE
-
Llama 3 Instruction Filtering: Quality and Complexity Evaluation
By
–
Llama 3's paper is full of cool insights. Here's how they filtered out bad instruction samples. – Quality is evaluated by a reward model and an LLM-as-a-judge
– Complexity is a combination of Instag (
https://
arxiv.org/abs/2308.07074) and LLM-as-a-judge + It looks like models are merged -
Zuckerberg Deep Dive: Llama 3.1 405B Open Source Future
By
–
Read our exclusive deep dive with Mark Zuckerberg, diving into Llama 3.1 405b. We cover: -All the details for Llama 3.1 405B
-Why open source for the greater good
-Meta’s lessons from the past
-Zuck’s prediction for the future of AI agents
-& more -
Meta’s Llama 3.1 Open-Source Release and AI Regulation Concerns
By
–
Thank you Meta and the Llama team for your huge contributions to open-source! Llama 3.1 with increased context length and improved capabilities is a wonderful gift to everyone. I hope foolish regulations don't like California's proposed SB1047 don't stop such innovations.
-
Llama 3.1 Launch on Google Cloud Today
By
–
Today's the day! Excited to see what people build with Llama 3.1 on Google Cloud!
-
Llama 3.1 Launch: New Language Model Family Disrupts Market
By
–
LLAMA 3.1 – ¡DIRECTO INMINENTE! Preparaos un café fresquito porque hoy vamos a desgranar una de las noticias del mes. La salida de una nueva familia de Llamas que ponen, una vez más, patas arriba el mercado de los modelos de lenguaje 🙂
-
Clone and Fork Llama Chat App for Custom Versions
By
–
Clone and fork this app to create your own Llama Chat!
-
Chat with Meta’s New Llama 3.1 405b Frontier Model
By
–
Chat with Llama 3.1 405b, the new frontier model from Meta. http://
llama3.replicate.dev -
Llama Models: Synthetic Data, RLHF, Agents Toward Open AGI
By
–
pod with @ThomasScialom of @AIatMeta
! Llama 2, 3 & 4: Synthetic Data, RLHF, Agents on the path to Open Source AGI https://
latent.space/p/llama-3 shoutouts:
– Why @ylecun
's Galactica Instruct would have solved @giffmana
's Citations Generator
– Beyond Chinchilla-Optimal: 100x