Show HN: OpenSciEval-AI Deriving Prime Theorem from Chaos (github.com)

A new AI evaluation framework called OpenSciEval (OSE) has been introduced to enhance how the AI community measures scientific creativity. Traditional AI assessments focus on knowledge recall, often overlooking the innovative capabilities required to forge novel scientific concepts. The inaugural "Prime–Chaos Challenge" showcased advanced models such as Gemini and Qwen, which successfully identified complex connections between prime number theory and chaotic dynamical systems, further promoting the use of physics-inspired methods in mathematical interpretation. This represents a notable shift in evaluating AI from problem-solving to understanding and generating scientific theories. The OSE framework facilitates a universal benchmark for assessing AI's creativity across various disciplines, focusing on the transformation of static mathematical processes into dynamic models. In testing, the Gemini 3 Pro model demonstrated significant potential by constructing a comprehensive framework for prime number chaotic dynamics, exhibiting capabilities in experimental design and innovation. The results reveal that using deterministic chaotic systems can better simulate prime distribution characteristics compared to traditional stochastic models, presenting critical insights for both AI development and mathematical inquiry. This initiative heralds a new direction in AI research, emphasizing the importance of creativity in scientific exploration.