LLMs are plateauing and the gap between closed vs. open is almost closed!!

If you are look at MMLU open-source is caught up to closed source and we are seeing the LLMs plateau

It's time to move on to different benchmarks that measure LLM capabilities on hard problems

The key

This is the primary problem I've experienced with gpt4o as well

Inverse Reinforcement Learning with Natural Language Goals

arxiv.org/pdf/2008.06924

How Self-Supervised Reinforcement Learning Combined With Offline Reinforcement Learning (RL) Could Enable Scalable Representation Learning

marktechpost.com/2021/12/19/uc-…

are you really ready for Mobius people? example image:

blog.glyph.im/2024/05/grand-…
Knut Jägersberg Chidambara .ML. Kevin Corella N. 🔬🔭 🇪🇺
Eveline Ruehlin Estela Mandela Dr. de l'immatériel
Mack #TechforGood Deus ex Machina BusinessIntelligence
Andrew Neff
HeidiLeeStockenstrom☮️ Fati Sule Dr. Monika Sonu I Founder Healthinnovationtoolbox
Dr. Khulood Almani | د.خلود المانع Prof. Sally Eaves Mike Flache

New hints about GPT-5 in OpenAI Vivatech Paris Talk

The hints are at:

18:00 where he shows the chart of intelligence of 'GPT-Next' arriving in 2024 - he talks about that model being a 'step function' over GPT-4 and 'become masters students in a blink of an eye',

and also

Excited to share our new paper, Lessons From The Trenches on Reproducible Evaluation of Language Models!

In it, we discuss common challenges we’ve faced evaluating LMs, and how our library the Evaluation Harness is designed to mitigate them 🧵

arxiv.org/abs/2405.14782

Do you know any tool/service to analyze/comment papers collectively with the proper AI features?

Why are leading technologists choosing Retrieval-Augmented Generation (RAG) systems for cutting-edge LLM solutions?

RAG connects LLMs with real-world data, tackling challenges like hallucinations and rising costs. Explore the top 5 reasons enterprises are choosing RAG systems

⚡️ Putin looking for ceasefire to cement gains in Ukraine, Reuters reports citing sources.

Russian President Vladimir Putin is open to a ceasefire that recognizes the current front lines on the battlefield but will fight on if Ukraine and its allies do not agree, Reuters

Check out the impact of AI-powered background verification on talent acquisition - bit.ly/3WYECzv

#AI #DataScience #HR #Technology #RecruitingAutomation #TalentAcquisition #AI #Hiring #HR Tech

cc: Tobias Kintzel Knut Jägersberg Tony Moroney Estela Mandela

Nous Research is hiring!

Apply Here: forms.gle/UWx2Pht8qioi1b…

So excited to join Applied Conformal Prediction 's course and learn from the best Valeriy M., PhD, MBA, CQF . maven.com/s/ae4546ceff

Très content d’avoir pu présenter Common Corpus avec Anastasia Stasenko à la journée Deep learning pour la Science au CNRS 🌍 Et au-delà la nécessité scientifique et éthique d’un tournant science ouverte pour l’entraînement de LLM.

Deploying LLMs: Top Learning & Educational Resources to Get Started by Youssef Hosni in gitconnected | Developer Community 💻 levelup.gitconnected.com/deploying-llms…
Knut Jägersberg Chidambara .ML. Kevin Corella N. 🔬🔭 🇪🇺
Eveline Ruehlin Estela Mandela Dr. de l'immatériel

It’s weird that big G trigger LLM
summarization for low volume queries and also allowing it to quote from low quality sources.

How to Optimize Chunk Sizes for RAG in Production? by i_ThinkBot pub.towardsai.net/how-to-optimiz…
Knut Jägersberg Chidambara .ML. Kevin Corella N. 🔬🔭 🇪🇺
Eveline Ruehlin Estela Mandela Dr. de l'immatériel