This repository distributes full model outputs for the BabyLM 2023 evaluation pipeline as well as scores per task and subtask for a subset of the submissions to the 2023 BabyLM Challenge. All outputs and scores are scraped from DynaBench. We additionally include links to published models.
Currently, we have uploaded the following content:
paper_summary.csv: Metadata about all papers submitted to the challenge.model_summary.csv: Metadata and aggregate task scores from all models submitted to dynabench.model_summary.xlsx: The above 2 datasets in excel format (best viewed in Google Sheets).dynabench_results.zip: The results for each subtask for each model submitted to dynabench. This.zipcontains one file for each track.
** Remaining contents in progress **