AIME-26

The AIME-26 dataset/benchmark contains unlabeled seed questions across math, science, and coding domains, and is used to evaluate the self-improvement capabilities of post-trained large language models through self-curated data generation and validation.

Papers using AIME-26 (2)

Self-Verified Distillation: Your Language Model Is Secretly Its Own Synthetic Data Pipeline2026VibeThinker-3B: Exploring the Frontier of Verifiable Reasoning in Small Language Models2026