self-consistency
Sampling several reasoning paths for the same question and taking the most common final answer.
In more detail
Wang and colleagues showed in 2022 that voting over chain-of-thought samples beats a single greedy answer on reasoning tasks. It multiplies cost by the number of samples, and if most samples share a mistake, the vote picks it confidently.
Related terms
- chain of thought — Prompting a model to write out intermediate reasoning steps before its final answer.
- sampling (decoding) — Choosing each next token at random from the model's probability distribution, shaped by settings such as temperature, top k and top p.
Taught in
- Reasoning Techniques — AI Engineering: LLM Apps, RAG & Agents
- Workflows Before Agents — AI Engineering: LLM Apps, RAG & Agents