See how Wayfair improved model accuracy by adopting data-centric AI to curate high-quality training data: Label millions of images in months vs. years Improve F1 scores by 20+ percentage points Deliver production models 10x faster Register: https://
buff.ly/4aXqt9w
MACHINE LEARNING
-
Wayfair Improves Model Accuracy with Data-Centric AI
By
–
-
Few-shot learning enhances fine-tuned model performance
By
–
Never underestimate the power of few-shot on top of fine-tuned models
-
Improve Fine-Tuned Models With Few-Shot Data Generation
By
–
High alpha trick to improve your fine-tuned models: Step 1: fine-tune your model on the data you have, do RLHF, whatever you want to do
Step 2: choose some of the best examples you have, and few-shot your model with those examples to generate more
Step 3: train on those! -
Data Rephrasing, Multiple Epochs, and Research Funding Priorities
By
–
– data rephrasing https://
arxiv.org/abs/2401.16380 – multiple epochs https://
latent.space/p/jan-2024 – give @BlancheMinerva more money and grad students -
Understanding How Large Language Models Actually Work
By
–
It's not an unreasonable mental model to form to be honest – enormous weird blobs of vector floating point matrices and token embeddings are hardly an obvious way that this technology might work
-
Fine-tuning vs RAG: Teaching AI Models Effectively
By
–
This could also explain why so many people instantly assume that "fine-tuning" a model is the obvious right way to teach it new information, as opposed to using more effective but less obvious techniques like RAG (not helped by that being a pretty terrible acronym)
-
Common Misconceptions About How AI Models Learn
By
–
I wonder how common it is for people to confidently hold an inaccurate idea of how AI models work where they believe that anything they show the model is instantly memorized and added to its "knowledge" of the world
-
How Models Actually Work: Training Misconceptions
By
–
I'm beginning to suspect that many people – including many corporate decision makers – have a confident but inaccurate model of how models work, where they think that absolutely anything the model sees is "training" which it then perfectly memorizes and adds to its knowledge
-
LLMs Unnecessary When Cheaper Models Perform Better
By
–
You absolutely don't need LLMs for this. Using LLMs here where a model that's 0.1% as expensive would work 5x better would be a show of gross incompetence.