Whether you're at #EMNLP2024 in person or following from your feed, here are research papers being presented by AI research teams at Meta to add to your reading list. Distilling System 2 into System 1: https://
go.fb.me/5l9832 Altogether: Image Captioning via Re-aligning
RESEARCH
-
Meta Research Papers at EMNLP 2024 Conference
By
–
-
Tribute to George Bekey, Pioneer Roboticist and Mentor
By
–
Just learned that George Bekey, one of my mentors @USC
, passed away in October. In 1993 we co-edited a book titled "Neural Networks in Robotics". He was an O.G. roboticist and a warm human being. RIP George. -
Groq ML Engineer Presents at Supercomputing Conference November 18
By
–
Don't miss this talk by Sanjif Shanmugavelu, Groq ML Engineer, at @Supercomputing on Monday, Nov 18. https://
sc24.conference-program.com/presentation/?
id=ws_corr108&sess=sess759
… -
Ilya’s Vision: Why He Built ASI and Our AI Future
By
–
What did Ilya see and why did he go building ASI?
— Chubby♨️ (@kimmonismus) 14 novembre 2024
Obviouslyhe knows more than any of us. However, he seems so certain that we are at the cusp of something bigger than we could have ever hoped for.
"Playing God, scientists have been accused of playing God for a while, but there… pic.twitter.com/YcWB2LB9CqWhat did Ilya see and why did he go building ASI? Obviouslyhe knows more than any of us. However, he seems so certain that we are at the cusp of something bigger than we could have ever hoped for. "Playing God, scientists have been accused of playing God for a while, but there
-
Nexusflow Releases Athene v2 72B Competitive Model
By
–
Nexusflow released Athene v2 72B – competetive with GPT4o & Llama 3.1 405B Chat, Code and Math > Arena Hard: GPT4o (84.9) vs Athene v2 (77.9) vs L3.1 405B (69.3) > Bigcode-Bench Hard: GPT4o (30.8) vs Athene v2 (31.4) vs L3.1 405B (26.4) > MATH: GPT4o (76.6) vs Athene v2
-
Small Language Models Era: Efficiency and Accuracy at Scale
By
–
We're in our #small model era And, it's not just us. If you look at any organization seriously deploying GenAI at #scale, you'll find that they are adopting small specialized models (#SLMs) to unlock better efficiency and #accuracy. Check out our latest discussion with
-
AI Companies Explore New Techniques to Advance LLM Performance
By
–
How AI companies are exploring new techniques to keep advancing model performance Via The Information As improvements to LLMs, AI companies are exploring a mix of established and innovative techniques to keep advancing model performance. Here’s a look at the strategies they’re
-
AI Revolution: Latest Trends Transforming Industries
By
–
Exciting times in the world of AI! Join me in my latest video as we dive into the revolutionary ways artificial intelligence is transforming industries. From real-world applications to the future of automation, I'll guide you through the latest AI trends that are reshaping our
-
AI Model Training: Settings Impact on Performance Time
By
–
once you're set with settings, click train!
— KREA AI (@krea_ai) 14 novembre 2024
the training will start and a percentage will appear showing progress.
the time it takes for a training to finish will depend on the settings you chose on the previous steps – large batch size and training steps make things slow. pic.twitter.com/tnTba6Z7llonce you're set with settings, click train! the training will start and a percentage will appear showing progress. the time it takes for a training to finish will depend on the settings you chose on the previous steps – large batch size and training steps make things slow.
-
Optimizing AI Model Training: Essential Settings and Parameters
By
–
one of the most important aspects of the trainer are the training settings.
— KREA AI (@krea_ai) 14 novembre 2024
as a rule of thumb:
– the more images he more training steps you need.
– if a training feels like regular flux, higher learning rate.
– high batch size can stabilize training but requires more steps.
-… pic.twitter.com/ORR41lp5f5one of the most important aspects of the trainer are the training settings. as a rule of thumb: – the more images he more training steps you need.
– if a training feels like regular flux, higher learning rate.
– high batch size can stabilize training but requires more steps.
–