What is SwiGLU and in which models is this activation used?
Machine Learning / AI
What are adapter tuning and prefix tuning?
What is a token in BERT and what tokenization options exist?
What is pose estimation (top-down vs bottom-up)? What is OpenPose?
How is context considered in modern models?
What task does GPT solve?
What are labeling functions?
What is homomorphic encryption and why is it used in ML?
What is ALiBi and how does it implement context length extrapolation?
What is OpenAI Assistants API and what are its limitations?
How does BERT predict masked words and consider the context of neighboring words?
What are typical bottlenecks in LLM serving?
What is self-consistency CoT?
What is judge LLM and what are its bias issues (length bias, position bias)?
How does the attention mechanism work in BERT and why is it more effective than Word2Vec?
What is repetition penalty and frequency/presence penalty?
What is query understanding and what are its tasks?
Tell us about your work experience at Sberbank?
What is Arrow and where is it used?
How does cosine distance help measure the semantic similarity of vectors?