MetaLead: A Comprehensive Human-Curated Leaderboard Dataset for Transparent Reporting of Machine Learning Experiments
This addresses the need for better benchmarking tools in ML research, though it is incremental as it builds on existing automation efforts by adding human curation and richer metadata.
The authors tackled the problem of limited and opaque leaderboard datasets in machine learning by creating MetaLead, a human-annotated dataset that captures all experimental results and includes extra metadata like experimental types and dataset splits, enabling more transparent and nuanced evaluations.
Leaderboards are crucial in the machine learning (ML) domain for benchmarking and tracking progress. However, creating leaderboards traditionally demands significant manual effort. In recent years, efforts have been made to automate leaderboard generation, but existing datasets for this purpose are limited by capturing only the best results from each paper and limited metadata. We present MetaLead, a fully human-annotated ML Leaderboard dataset that captures all experimental results for result transparency and contains extra metadata, such as the result experimental type: baseline, proposed method, or variation of proposed method for experiment-type guided comparisons, and explicitly separates train and test dataset for cross-domain assessment. This enriched structure makes MetaLead a powerful resource for more transparent and nuanced evaluations across ML research.