Shading and Lighting Models in Computer Graphics

This review paper covers 30 important research papers published between 2021 and 2025 on the topic of shading and lighting models in computer graphics. These five years were one of the most exciting and fast-moving periods the field has ever seen, driven largely by the explosion of neural rendering techniques and the rapid advancement of real-time ray tracing hardware. The papers reviewed here cover a wide range of topics. On the neural side, we look at Neural Radiance Fields (NeRF) and their descendants, which introduced a completely new way of representing and rendering 3D scenes using neural networks. We review how these methods were extended to handle material decomposition, relighting, and reflections. We also look at 3D Gaussian Splatting, a newer technique that achieves real-time rendering speeds while maintaining high visual quality. On the real-time rendering side, we examine major advances in global illumination including ReSTIR GI, Lumen (Unreal Engine 5), and radiance caching systems. We also review neural denoising — the use of deep learning to clean up the noise produced by path tracing — which has become an essential part of every modern real-time renderer. Each paper is summarized in plain, accessible language. For every paper, we describe what problem it solved, how the researchers approached the solution, and what impact the work had. No deep technical background is required to understand the summaries. Together, these 30 papers paint a picture of a field in rapid transformation, moving from manually engineered rendering algorithms toward hybrid systems that combine physics-based rendering with powerful machine learning.