NeurIPS 2025 E2LM Competition: Call for Contributions

Participate in the NeurIPS 2025 E2LM Competition to develop benchmarks for early-stage LLM evaluation. Register now and compete for prizes.

The NeurIPS 2025 E2LM Competition invites participants to develop benchmarks that capture early-stage reasoning and scientific knowledge in Large Language Models (LLMs).

Existing evaluation benchmarks often fail to provide meaningful signals during the initial stages of LLM training. This competition aims to build new benchmarks to effectively capture relevant signals in early training stages, specifically for the scientific knowledge domain.

• Signal Quality: smooth, meaningful learning curves
• Ranking Consistency: stable model rankings across training
• Scientific Compliance: benchmarks should accurately reflect scientific knowledge and reasoning

To participate, register at https://e2lmc.github.io/registration and submit solutions through a HuggingFace Space. A comprehensive starting kit is available at https://e2lmc.github.io/starter_kit.

Prizes include $6,000 for 1st place, $4,000 for 2nd place, and $2,000 for 3rd place, with additional awards for best student submissions.

Tags: NeurIPS 2025, E2LM Competition, Large Language Models, LLMs, Benchmarks, AI Research, NLP