uqlm.utils.prompts.decomposition.get_mohri_2024_breakdown_prompt

uqlm.utils.prompts.decomposition.get_mohri_2024_breakdown_prompt(response)

Claim decomposition prompt adapted from Mohri & Hashimoto (2024).

Instructs the LLM to break input into small, independent, non-overlapping claims. The original paper used JSONL output with confidence scores; this version is adapted to use the ### prefix format for compatibility with uqlm’s claim parser.

Return type:

str

Parameters:

response (str) – The response to be broken down into independent claims.

Returns:

The prompt template for breaking down the response into non-overlapping claims.

Return type:

str