Self-Consistency
A widely used inference-time technique that improves the reasoning accuracy of language models by sampling multiple reasoning paths for the same question and taking a majority vote over the final answers.
Timeline (last 90 days)
The PSV framework unifies self-consistency and verifier-guided selection: perturbation generation plays the sampling role, validation-set selection corresponds to verifier screening, and vote aggregation is consistent with self-consistency.
Multi-prompt aggregation is similar to self-consistency sampling, but the difference is that the dimension of variation is the prompt rather than decoding randomness.
All Facts (2)
The PSV framework unifies self-consistency and verifier-guided selection: perturbation generation plays the sampling role, validation-set selection corresponds to verifier screening, and vote aggregation is consistent with self-consistency.
50%UnverifiedMulti-prompt aggregation is similar to self-consistency sampling, but the difference is that the dimension of variation is the prompt rather than decoding randomness.
50%