Direct-Inverse Prompting: Analyzing LLMs' Discriminative Capacity in Self-Improving Generation
This work addresses a key reliability issue for users of LLMs by providing a method to reduce output variability, though it is incremental as it builds on existing discriminative capabilities.
The paper tackles the problem of generative uncertainty in LLMs, where outputs vary despite similar inputs, by proposing and analyzing three discriminative prompts (direct, inverse, hybrid) to help LLMs self-improve in identifying correct answers, with results showing which prompt works best under specific conditions.
Mainstream LLM research has primarily focused on enhancing their generative capabilities. However, even the most advanced LLMs experience uncertainty in their outputs, often producing varied results on different runs or when faced with minor changes in input, despite no substantial change in content. Given multiple responses from the same LLM to the same input, we advocate leveraging the LLMs' discriminative capability to reduce this generative uncertainty, aiding in identifying the correct answers. Specifically, we propose and analyze three discriminative prompts: direct, inverse, and hybrid, to explore the potential of both closed-source and open-source LLMs in self-improving their generative performance on two benchmark datasets. Our insights reveal which discriminative prompt is most promising and when to use it. To our knowledge, this is the first work to systematically analyze LLMs' discriminative capacity to address generative uncertainty.