Self-Consistency Sampling
A test-time compute enhancement technique that generates multiple reasoning paths for the same question and selects the most consistent answer through majority voting, thereby improving the reasoning accuracy of large models.
Core Facts
The core idea of the test-time compute paradigm is to shift resources from the training phase to the answer generation process, trading more inference compute for higher-quality output.
Test-Time Compute refers to the mechanism by which new-generation models perform longer internal reasoning via chain-of-thought before generating a final answer.
Timeline (last 90 days)
The average logprob over an answer sequence can serve as a proxy for overall confidence, and self-consistency sampling assesses confidence by observing answer agreement across multiple samples.
All Facts (3)
The core idea of the test-time compute paradigm is to shift resources from the training phase to the answer generation process, trading more inference compute for higher-quality output.
75%VerifiedTest-Time Compute refers to the mechanism by which new-generation models perform longer internal reasoning via chain-of-thought before generating a final answer.
65%UnverifiedThe average logprob over an answer sequence can serve as a proxy for overall confidence, and self-consistency sampling assesses confidence by observing answer agreement across multiple samples.
50%