What is image captioning and what metrics evaluate it (CIDEr)?
Machine Learning / AI
What architectural trends are present in modern decoder-only LLMs (LLaMA, Mistral, Qwen)?
What is belief state and how is it updated?
How to handle rare hieroglyphic characters and emojis in the tokenizer?
What is negative sampling in training word2vec and why is it needed?
What are the practical problems of RLHF (reward hacking, mode collapse)?
What are the safety features when managing ML-controlled robots?
What is Prophet and what are its strengths and weaknesses?
Why is cosine similarity often preferred over Euclidean distance for semantic search?
What are the modern trends in training on mixed GPU+CPU+TPU clusters?
Tell us about bag-of-words and its limitations.
What are the features of active learning in computer vision?
What is CopyPaste augmentation?