Answer: (b)

Explanation:
Large Language Models work by learning statistical patterns in language data.

Statement 1 is correct. LLMs generate text by estimating probabilities for the next token (word/sub-word) and selecting the most likely one (or sampling from likely options). This is the core mechanism behind autoregressive language generation.

Statement 2 is correct. During training, LLMs use mathematical optimization techniques (like gradient descent) to minimize prediction error, typically through a loss function such as cross-entropy loss.

Statement 3 is incorrect. LLMs do not produce inherently unbiased outputs. They can reflect biases present in training data and model design, which is a well-known limitation in machine learning systems.