Quiz: Best Practices in Handling and Preprocessing User Input for LLMs — Questions & Answers

This quiz contains 10 questions. Below is a complete reference of all questions, answer choices, and correct answers. You can use this section to review after taking the interactive quiz above.

1. Question 1: Why is chunking user input before submitting to an LLM important?

   Why should unstructured user data, such as text from forms or chat logs, be chunked before submitting it to a large language model?

   • It enhances retrieval quality and reduces hallucination risk
   • It increases computation time and model error
   • It merges unrelated data into one block
   • It always reduces the accuracy of responses
   • It ensures that the LLM receives only numeric input
   Show correct answer

   Correct answer: It enhances retrieval quality and reduces hallucination risk

2. Question 2: Choosing Effective Chunk Size

   Which principle should guide the selection of chunk size when splitting user input for an LLM?

   • Balance between including enough context without exceeding the model's context window
   • Always use the smallest possible chunk (single word)
   • Chunks should always be at least 1000 characters
   • Larger chunks always produce more accurate answers
   • Chunking does not affect LLM performance
   Show correct answer

   Correct answer: Balance between including enough context without exceeding the model's context window

3. Question 3: Overlap in Chunking

   What is the main purpose of introducing overlapping sections between chunks when preparing text for LLM input?

   • To preserve semantic context that may span across chunk boundaries
   • To increase redundancy and storage usage unnecessarily
   • To make chunking more random
   • To reduce the number of relevant chunks retrieved
   • To speed up the chunking process with no other benefit
   Show correct answer

   Correct answer: To preserve semantic context that may span across chunk boundaries

4. Question 4: Semantic Chunking

   What does semantic chunking refer to when processing unstructured user text for LLMs?

   • Splitting text based on meaningful units such as headings and sections
   • Cutting text into fixed-size byte blocks only
   • Merging all data into a single chunk
   • Chunking at arbitrary word intervals
   • Ignoring data contained in tables and images
   Show correct answer

   Correct answer: Splitting text based on meaningful units such as headings and sections

5. Question 5: Prompt Engineering for Extraction

   Which practice enhances the reliability of LLM data extraction from user input?

   • Providing clear instructions and optionally few-shot examples in prompts
   • Adding unrelated questions to confuse the model
   • Using only system prompts with no user data
   • Avoiding examples in the prompt at all costs
   • Submitting raw, unformatted input without prompting
   Show correct answer

   Correct answer: Providing clear instructions and optionally few-shot examples in prompts

6. Question 6: Use of Constrained Decoding

   How can constrained decoding make LLM output handling user input safer and more structured?

   • By limiting the generation to valid output formats and values
   • By encouraging the model to hallucinate new data
   • By forcing the LLM to ignore input details
   • By generating outputs only in natural language
   • By randomly skipping parts of the text
   Show correct answer

   Correct answer: By limiting the generation to valid output formats and values

7. Question 7: Handling Multimodal Input

   When should a vision-language model (VLM) be used instead of a text-only LLM for preprocessing user data?

   • When the user data includes rich images, tables, or non-machine-readable depictions
   • Only when the data is exclusively in plain text paragraphs
   • When audio files need to be transcribed
   • When input data is all in XML format
   • Whenever lower computation cost is the priority
   Show correct answer

   Correct answer: When the user data includes rich images, tables, or non-machine-readable depictions

8. Question 8: Evaluating Extraction Performance

   Which metrics are most relevant for evaluating the effectiveness of structured data extraction from user input with LLMs?

   • Precision, recall, and F1-score based on TP, FP, and FN entries
   • Pixel accuracy and frame rate
   • Page count and file size
   • Number of chunks per request
   • Training time of the underlying model
   Show correct answer

   Correct answer: Precision, recall, and F1-score based on TP, FP, and FN entries

9. Question 9: Handling Complex Document Structures

   How should unstructured documents containing tables, plots, and unusual formatting be handled before submitting to LLMs?

   • Preprocess with specialized tools to partition and extract structured elements
   • Ignore non-text elements and only submit paragraph text
   • Merge all data into a single text chunk regardless of type
   • Convert all content to images
   • Only process the text that directly follows tables and images
   Show correct answer

   Correct answer: Preprocess with specialized tools to partition and extract structured elements

10. Question 10: Addressing Ambiguities in Text

    What is a recommended approach for splitting user input when pronouns or references may cross sentence boundaries?

    • Use chunking methods that incorporate sentence overlap or semantic similarity
    • Always split after every period, regardless of context
    • Do not chunk, send full text as-is
    • Split strictly by fixed character count only
    • Chunk at random intervals to diversify input
    Show correct answer

    Correct answer: Use chunking methods that incorporate sentence overlap or semantic similarity