Google

The search phase in LLM inference involves finding an output sequence `y` that is considered optimal or sub-optimal. The quality of any candidate sequence is determined by its conditional log-probability, `log Pr(y|x)`, which the search process aims to maximize.

Search for Optimal or Sub-optimal Sequences in LLM Inference

A language model is tasked with completing the sentence: "The inventor of the light bulb was...". The search process, designed to find the sequence with the highest possible conditional log-probability, generates the following output: "The inventor of the light bulb was was was was was...". Explain why a search algorithm strictly aiming to maximize the sequence's probability might produce such a repetitive and unhelpful result, even if it is mathematically "optimal" according to the model's calculations.

Limitations of Probability Maximization in Text Generation

An LLM is generating a completion for the prompt: 'To improve team collaboration, we should...'. The inference process has produced three candidate sequences, and the model has calculated the total conditional log-probability for each one. Based on the objective of finding the most probable output sequence, which of the following options should be selected? Justify your choice.

Evaluating Candidate Sequences in LLM Inference

A language model is generating text with the goal of producing the most probable output sequence. It has already generated the phrase 'The best way to learn is by...' and must now decide the next word. The model calculates the following probabilities for the next possible words:

*   `Pr('doing' | 'The best way to learn is by...') = 0.6`
*   `Pr('reading' | 'The best way to learn is by...') = 0.3`
*   `Pr('sleeping' | 'The best way to learn is by...') = 0.09`
*   `Pr('car' | 'The best way to lea

Learn Before

Related