How does tokenization work? Why are subword tokens used instead of whole words?
Middle+
219
Machine Learning / AI
3611
0
Tell us about BGE-M3: three representations — dense, sparse, multi-vector.
Middle+
185
What are Recall and Precision? Task: 50 fruits, 30 apples, and 20 pears, a robot correctly identified 20 apples, but 15 pears were misclassified as apples. Calculate the metrics.
Middle+
174
Are you familiar with LoRA and fine-tuning methods?
Middle+
172
How were the coefficients for hybrid search selected?
Middle+
166
Why did you need 12,000 question-answer pairs when there were only 15 templates?
Middle+
151
Why did you recognize documents as images instead of creating a parser?
Middle+
148
How did you form the test data set — from the same 12 thousand pairs or from new data?
Middle+
143
Tell us about clustering: what problem did you solve, what method did you use, and how did you measure quality?
Middle+
140
How did you evaluate the quality of the RAG system? What metrics did you use?
Middle+
139
What is a dense vector (dense embedding)? How is it obtained? What is its size?
Middle+
138
How does a transformer work? Tell me about BERT and GPT.
Middle+
138
What models were used at each stage? For entity extraction, vectorization, generation?
Middle+
136
Why is it necessary to reduce the size of the dictionary? What are the disadvantages of a large dictionary? In which function does the load occur?
Middle+
136
What is a cross-encoder and where have you used it?
Middle+
135
What is a sparse vector? How does it differ from a dense one?
Middle+
134
What was sent to the end user in your system?
Middle+
134
What is the difference between GET and POST requests?
Middle+
130
Is it one of 12,000 answers or multiple answers suitable for a single query?
Middle+
128
Explain simply how your system works for a single user without asynchronous processes.
Middle+
127