The Parallel-R1 framework uses reinforcement learning to teach models how to explore multiple reasoning paths at once, ...

In everyday life and across nearly every industry, mathematical reasoning is becoming more essential. We need to rapidly expand access to the after-school and summer programs that help young people ...

The artificial intelligence community celebrated a remarkable milestone in 2025 when both Google DeepMind and OpenAI systems ...

The trend of AI researchers developing new, small open source generative models that outperform far larger, proprietary peers continued this week with yet another staggering advancement. Alexia ...

Students with strong analytical thinking, intellectual curiosity, and an interest in solving societal or business problems ...

Experiments show that Parallel-R1 not only brings an average accuracy improvement of up to 8.4% across multiple mathematical benchmarks but also achieves a 42.9% performance leap in the AIME25 test ...

The focus on Indian knowledge systems in the UGC’s proposed mathematics curriculum is better suited to other disciplines such ...

DeepSeek-R1 takes a different path by adopting a pure reinforcement learning framework and introducing the Group Relative Policy Optimization (GRPO) algorithm. During the training process, the model ...

ChatGPT shocked researchers by solving Plato’s ancient puzzle in a new way, showing reasoning-like behavior when guided with ...

Strong spatial skills are critical for everyday tasks and across many careers—they also strengthen students’ math performance ...

We've wondered for centuries whether knowledge is latent and innate or learned and grasped through experience, and a new ...