Large Language Models (LLMs) often produce outputs that - though plausible - can lack consistency and reliability, particularly in ambiguous or complex scenarios. Challenges arise from ensuring that outputs align with both factual correctness and human intent. This is problematic in existing approaches that trade improved consistency for lower accuracy. To mitigate these challenges, we propose a novel game-theoretic approach to enhance consistency and reliability during the decoding stage of LLM output generation. This ensures consistency through Correctness Alignment and enhances reliability via Ambiguity Calibration. Remarkably, our game design allows smaller models to outperform much larger models through game mechanisms (e.g. Large Language Models (LLMs) have demonstrated extraordinary capabilities in tasks such as factual question answering, fact-checking, and open-ended text generation (Brown et al., 2020; Radford et al., 2021). However, as these generative models increase in ...

Artificial intelligence research startup OpenAI on Tuesday introduced a tool that's designed to figure out if text is human-generated or written by a computer. The release comes two months after OpenAI captured the public's attention when it introduced ChatGPT, a chatbot that generates text that might seem to have been written by a person in response to a person's prompt. Following the wave of attention, last week Microsoft announced a multibillion-dollar investment in OpenAI and said it would incorporate the startup's AI models into its products for consumers and businesses. Schools were quick to limit ChatGPT's use over concerns the software could hurt learning. Sam Altman, OpenAI's CEO, said education has changed in the past after technology such as calculators has emerged, but he also said there could be ways for the company to help teachers spot text written by AI.