Reproducibility in NLP: What Have We Learned from the Checklist?
This study addresses reproducibility issues in NLP research by evaluating the impact of a checklist, offering insights for improving scientific practices in the field, though it is incremental as it builds on existing tools.
The paper analyzed 10,405 responses to the NLP Reproducibility Checklist, finding that its introduction increased reporting of key information like efficiency and hyperparameters, and submissions with more 'Yes' responses had higher acceptance rates, while only 46% open-sourced their code despite an 8% higher reproducibility score for those that did.
Scientific progress in NLP rests on the reproducibility of researchers' claims. The *CL conferences created the NLP Reproducibility Checklist in 2020 to be completed by authors at submission to remind them of key information to include. We provide the first analysis of the Checklist by examining 10,405 anonymous responses to it. First, we find evidence of an increase in reporting of information on efficiency, validation performance, summary statistics, and hyperparameters after the Checklist's introduction. Further, we show acceptance rate grows for submissions with more Yes responses. We find that the 44% of submissions that gather new data are 5% less likely to be accepted than those that did not; the average reviewer-rated reproducibility of these submissions is also 2% lower relative to the rest. We find that only 46% of submissions claim to open-source their code, though submissions that do have 8% higher reproducibility score relative to those that do not, the most for any item. We discuss what can be inferred about the state of reproducibility in NLP, and provide a set of recommendations for future conferences, including: a) allowing submitting code and appendices one week after the deadline, and b) measuring dataset reproducibility by a checklist of data collection practices.