verbalized-sampling

Generate diverse LLM outputs by prompting for a probability distribution.

95|8|Updated Mar 23, 2026

Search Tags:#brainstorming #LLM #prompting #diversity #verbalized-sampling #open-ended QA

Ready to run

npx skills add https://github.com/gnurio/nurijanian-skills --skill verbalized-sampling

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: verbalized-sampling
Source: https://github.com/gnurio/nurijanian-skills/tree/main/skills/verbalized-sampling
Command: npx skills add https://github.com/gnurio/nurijanian-skills --skill verbalized-sampling

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Verbalized Sampling (VS) addresses mode collapse in LLM outputs by requesting a probability distribution over multiple responses, increasing diversity and coverage in creative tasks, open-ended QA, and prototyping scenarios.

Core Features & Use Cases

Generate multiple candidate outputs with explicit probabilities for each result
Applies across creative writing, brainstorming, persona simulation, synthetic data generation, adversarial examples, and open-ended QA
Helps avoid the same-old cluster of ideas by encouraging tail and distribution-aware prompts

Quick Start

Prompt the model to produce k diverse outputs, each with a probability, and return them as a JSON array.

Frequently Asked Questions about verbalized-sampling

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I get diverse LLM outputs instead of the same repeated ideas when brainstorming?▼

To get diverse LLM outputs, prompt the model to generate a probability distribution over multiple responses rather than a single answer. This technique, Verbalized Sampling, counteracts mode collapse and ensures genuine diversity in brainstorming and creative writing.

Why does my LLM keep returning the same cluster of answers when I ask for multiple ideas?▼

LLMs return the same cluster of answers due to mode collapse caused by typicality bias in alignment data. Verbalized Sampling addresses this by requesting a probability distribution across responses, encouraging the model to explore tail and distribution-aware outputs.

Does verbalized sampling require model fine-tuning to generate diverse prompts?▼

Verbalized sampling requires no fine-tuning. It is a training-free technique that prompts the model to produce k diverse outputs with explicit probabilities, returning them as a JSON array to increase coverage in open-ended QA and synthetic data generation.

When should I use probability distribution prompting for open-ended QA?▼

Use probability distribution prompting for open-ended QA when the task needs genuine diversity and multiple valid answers exist. It is also effective for persona simulation, adversarial examples, and synthetic data generation where typicality bias limits output coverage.

How to generate synthetic data with diverse outputs using verbalized sampling?▼

To generate synthetic data with diverse outputs, prompt the model to produce k distinct candidate responses, each with an explicit probability. This verbalized sampling method formats results as a JSON array, avoiding typicality bias and expanding data coverage.