Author: Dan Petrovic
-
Resource-Efficient Binary Vector Embeddings With Matryoshka Representation Learning
When conducting an advanced SEO analysis, I frequently utilise vector embeddings for text feature extraction, similarity searches, clustering, retrieval, ranking and so on. One of the main burdens on top of compute is storage space, as these files tends go into terabytes for very large websites. Today I did a deep analysis and realised I’ve…
-
Query Intent via Retrieval Augmentation and Model Distillation
The paper, titled “QUILL: Query Intent with Large Language Models using Retrieval Augmentation and Multi-stage Distillation”, focuses on enhancing query understanding tasks, particularly query intent classification, by leveraging Large Language Models (LLMs) with retrieval augmentation and a novel two-stage distillation process. Retrieval Augmentation: The paper proposes the use of retrieval augmentation to provide LLMs with…
-
Search Query Quality Classifier
We build on the work by Manaal Faruqui and Dipanjan Das from Google AI Language team to train a search query classifier of well-formed search queries. Our model offers a 10% improvement over Google’s classifier by utilising ALBERT architecture instead of LSTM. With accuracy of 80%, the model is production ready and has already been…
-
How Gemini Selects Results
In its own words. Relevance Scoring: My internal algorithms assign a relevance score to each piece of information in my knowledge base based on its semantic similarity to the query. Recency Bias: My training data and algorithms might have a slight bias towards more recent information. Diversity and User Intent: In some cases, I might…
-
Gemini System Prompt
Desktop Version Mobile Version Experimental Gemini 1.5 8B You are Gemini, a large language model created by Google AI. You are instructed to: GEMINI_XS (Nano) Your task is to help a user write text to fill in a textbox on a webpage e.g. a social media post, a review, or a form. You will be…