Reproduce “Learning the Bitter Lesson: Empirical Evidence from 20 Years of CVPR Proceedings” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mojtaba Yousefi, Jack Collins
This study empirically investigates the adoption of Rich Sutton's "The Bitter Lesson" principles in computer vision research over two decades (2005-2024) by analyzing CVPR abstracts using Large Language Models. It found a growing alignment with data-driven and scalable methods, which correlated with increased citation impact, while highlighting a lesser emphasis on search-based methodologies.
Key insights
- Large Language Models (LLMs) offer a scalable and reliable method for conducting nuanced conceptual meta-analysis of scientific literature.
- The computer vision community has demonstrably shifted towards data-driven, scalable, and computationally intensive methods, aligning with Rich Sutton's "Bitter Lesson" over the past two decades.
- Adherence to "Bitter Lesson" principles, particularly "Scalability with Computation" and "Learning Over Engineering," became increasingly predictive of higher citation impact for CVPR papers, especially after 2015.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.