Machine Learning / AI

Tell me about your most interesting project in the field of NLP and LLM, including architecture and solution evaluation.

How to evaluate RAG systems? What metrics were used?

An agent never invokes one of the three tools. How to investigate and solve this problem?

Can you name some metrics for classification tasks? Tell me about each of them.

Tell me about ROC-AUC: what is it, and how is it used?

What is an agent? What are the main components needed to create an agent? How is memory implemented?

Tell me about LoRA: what it is, how it is trained, and how it works during inference.

How to protect RAG from prompt injection and unwanted content?

Tell about the main architecture of the transformer, what are its main parts.

How to aggregate and analyze markup with overlap of 3 (three operators per question, binary answers)?

What does a clustering task look like and what algorithms do you know?

What loss function is used in classification tasks?

Tell about model distillation.

What is RAG and what are its main components?

Tell about types of positional encoding, especially about Rotary Positional Encoding (RoPE).

Tell us about the parameters of generation: temperature, Top-P, and Top-K.

What is the difference between cross-attention, masked attention, and self-attention?

We have a region where a tornado can occur. We are developing a binary classification model. Which metric do we pay more attention to — Precision or Recall?

Tell me, what is a classification task?

What is hybrid search in RAG?