Term
Self-Consistency
Self-Consistency is a prompting technique where the same question is answered multiple times with chain-of-thought, and the most frequent answer is picked as the final result.
Self-Consistency — explained in more detail
Self-Consistency is an extension of chain-of-thought prompting (Wang et al., 2022). Rather than relying on a single reasoning path, the model is asked to solve the same question multiple times — typically 5 to 40 times with a higher temperature so that different solution paths emerge. The final answers are then voted on; the most frequent one wins (majority vote). The assumption: correct reasoning paths converge, faulty ones scatter.
Example / Practical use
Particularly effective for math, logic and code tasks with a single correct answer. On benchmarks like GSM8K or AIME, self-consistency delivers double-digit accuracy gains over plain chain-of-thought in some settings. The downside: inference cost scales linearly with the number of samples — 20 runs mean 20× the tokens.
How it differs from similar concepts
Unlike Tree of Thoughts, the reasoning paths are generated independently rather than branched in a tree structure. Unlike Self-Refine, there is no iterative improvement — the model samples in parallel and votes.
Entdecke mehr
Effort level and deep thinking: two independent axes for AI tasks
Effort scales breadth, deep thinking scales depth. When each setting makes sense — with three clear examples and one rule of thumb.
GlossarChain-of-Thought
Chain-of-Thought (CoT) is a prompting technique that asks the model to spell out its reasoning in intermediate steps — boosting accuracy on multi-step tasks.
LexikonPrompt Security — Injection, Leaking, Guardrails
How prompt injection, prompt leaking, and jailbreaks work — and which defenses (guardrails, spotlighting, sanitization) actually help.