E2LM Competition

NeurIPS 2025 E2LM Competition: Early Training Evaluation of Language Models

TII UAE and collaborators are announcing the Early Training Evaluation of Language Models (E2LM) competition at NeurIPS 2025. The competition focuses on predicting or evaluating language model capabilities from early training checkpoints, addressing the challenge of forecasting final model performance without completing full training runs. This is relevant to evaluation methodology and training efficiency research in the AI/ML community.