Which of the following statements with regard to Large Language Models (LLMs) used in machine learning is/are correct ? 1. LLMs assign probabilities to the next possible words and then pick the one with the highest probability. 2. LLMs process data through mathematical optimization to minimise prediction errors. 3. LLMs produce unbiased outputs. Select the answer using the code given below :

Option (B) — 1 and 2 only

• Large Language Models (LLMs) like GPT work by predicting the next token/word in a sequence. They assign probability distributions over possible next words and typically select based on highest probability (or use sampling strategies).
• LLMs are trained through mathematical optimization processes (gradient descent, backpropagation) that minimize a loss function, which measures prediction errors between the model output and actual training data.
• LLMs are known to produce biased outputs because they are trained on large corpora of text data from the internet that contain inherent biases related to gender, race, culture, and other factors. Bias in training data leads to biased outputs.

• 1. LLMs assign probabilities to the next possible words and then pick the one with the highest probability: Correct. This is the core mechanism of autoregressive language models, which generate text by predicting the most likely next token.
• 2. LLMs process data through mathematical optimization to minimise prediction errors: Correct. During training, LLMs use optimization algorithms like stochastic gradient descent to minimize the cross-entropy loss (prediction error).
• 3. LLMs produce unbiased outputs: Incorrect. LLMs are well-documented to produce biased outputs reflecting biases present in their training data. This is a widely recognized limitation of current LLM technology.