Posits Solve the Light Leak Problem in Computer Graphics

This work explores how posit arithmetic can resolve precision-driven artifacts in computer graphics, specifically light leaks that occur during triangle-ray intersection tests in rendering. Unlike IEEE floating-point numbers, posits provide tapered precision and exact dot products via a more exact environment known as the quire, enabling more robust handling of small and large values within the same scene. Through comparative tests using 32 and 16 bit posits and floats, we demonstrate that posits significantly reduce false-positive intersections and visual artifacts in edge-aligned geometry. This highlights posits as a promising alternative for computer graphics pipelines where accuracy is critical.