Showing Your Work Doesn't Always Work
This work addresses statistical pitfalls in reporting practices for NLP researchers, offering an improved method to enhance result reliability.
The authors critically examined a popular method for reporting neural network experimental results, showing that the advocated estimator is biased and error-prone, and they derived an unbiased alternative with empirical evidence from simulations.
In natural language processing, a recently popular line of work explores how to best report the experimental results of neural networks. One exemplar publication, titled "Show Your Work: Improved Reporting of Experimental Results," advocates for reporting the expected validation effectiveness of the best-tuned model, with respect to the computational budget. In the present work, we critically examine this paper. As far as statistical generalizability is concerned, we find unspoken pitfalls and caveats with this approach. We analytically show that their estimator is biased and uses error-prone assumptions. We find that the estimator favors negative errors and yields poor bootstrapped confidence intervals. We derive an unbiased alternative and bolster our claims with empirical evidence from statistical simulation. Our codebase is at http://github.com/castorini/meanmax.