I also wonder if taking the mean of several embeddings makes them “out of distribution” somehow im really not sure, about the muster question! I convinced myself it’s because embeddings are sort of “compressed” and it takes a lot of effort to decompress them
MACHINE LEARNING
-
Generative AI Hype: New Trend for Clueless Adopters
By
–
'Generative AI' is the new 'Machine Learning' for those who jump on trends without a clue. We dont have data but we want to use machine learning. We don't know anything about machine learning but we will use it. Now the same is for gen ai ¯_(ツ)_/¯
-
LoRAX Open Source Framework Reaches 1000 Stars
By
–
LoRAX, our #opensource framework for cost efficiently serving many #finetuned LLMs on as single GPU, recently joined the 1k club! Check out #LoRA Land to see it in action: https://
pbase.ai/3SXxeAL Download #LoRAX and give it a star if you like it: https://
pbase.ai/49vp89y -
SAS Academy for Data Science: Elevate Your Skills March 21
By
–
Elevate your skills using SAS + keep pace with the most innovative tech. Join us on March 21 to learn more about the SAS Academy for Data Science. http://
2.sas.com/6012nYq6s #DataScientist -
1-Bit LLMs: New Era in Language Model Quantization
By
–
The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits Ma et al.: https://
arxiv.org/abs/2402.17764
v1
… #ArtificialIntelligence #DeepLearning #MachineLearning -
Do Frontier LLMs Actually Reason? New Evidence Suggests Yes
By
–
It is fascinating that, over a year after ChatGPT released, we still don't actually know when, if, and how well frontier LLMs "reason." There have been some papers suggesting limits to reasoning, but this new work finds that LLMs, while not as good as humans, can do reasoning. https://t.co/PlMiNE2RFJ
— Ethan Mollick (@emollick) 28 février 2024It is fascinating that, over a year after ChatGPT released, we still don't actually know when, if, and how well frontier LLMs "reason." There have been some papers suggesting limits to reasoning, but this new work finds that LLMs, while not as good as humans, can do reasoning.
-
Three Training Phases Required for LLM Performance
By
–
Professor Stephen Bach offers an exceptionally clear description of the three training phases required to make LLMs perform. https://
snorkel.ai/large-language
-model-training-three-phases-shape-llm-training/
… #largelanguagemodels #llmtraining #genai -
Critique of MediaTek’s low-frame-rate diffusion model demo
By
–
Il y a des démos qui laissent sur notre faim, on va dire… Ici, le stand de MediaTek a fait une démo avec un storytelling autour « d’un voyage dans le futur », où le dispositif était « juste » un modèle de diffusion guidé avec un input et une fréquence d’image de même pas 1 fps.… pic.twitter.com/B6NdDm34LO
— Defend Intelligence (Anis Ayari) (@DFintelligence) 28 février 2024Il y a des démos qui laissent sur notre faim, on va dire… Ici, le stand de MediaTek a fait une démo avec un storytelling autour « d’un voyage dans le futur », où le dispositif était « juste » un modèle de diffusion guidé avec un input et une fréquence d’image de même pas 1 fps.
