![alt](/assets/images/hero-237ed4b707a303e4ad3353daaf4edab8.jpeg)
Matryoshka Representation Learning
Overview of OpenAI's New Truncatable - Matryoshka Embeddings
January 29, 2024 路 2 min read
Overview of OpenAI's New Truncatable - Matryoshka Embeddings
A breakdown of the different training steps that go into creating a LLM.
Finetuning a 7B model to outperform GPT-4 for hallucination detection.
Provides a taxonomy of different types of hallucinations.
32k context length retreival models with sub-quadratic attention mechanism.
Using persuasion ot jailbreak LLM's.