NLP progress

For each task, it lists the datasets used for evaluation, the current best-performing models and their scores, and links to the papers that produced those resultsserving as a real-time leaderboard of NLP research progress.

The repository covers over 100 NLP tasks organized by category: text classification, sequence labeling, parsing, semantic tasks, language modeling, machine translation, speech tasks, and emerging areas like dialogue systems and commonsense reasoning.

This breadth makes it useful not just for tracking specific benchmarks but for understanding the full landscape of what NLP systems are measured on and what the gaps between current and human-level performance look like across different problem types.

NLP researchers identifying which tasks have been solved versus which remain open challenges for new research, practitioners evaluating which techniques and models are current best practice for a specific NLP task they're working on, and anyone trying to understand the overall trajectory of the NLP field use NLP Progress as a reference.

The collaborative GitHub maintenance model means entries are updated as new papers improve state-of-the-art results, though the pace of NLP advancement means some entries lag behind the latest published resultsa known limitation acknowledged by the maintainers.