Nvidia's Nemotron-Cascade 2 wins math and coding gold medals with 3B active parameters — and its post-training recipe is now open-source

Nvidia's Nemotron-Cascade 2 is a 30B MoE model that activates only 3B parameters at inference time, yet achieved gold medal-level performance at the 2025 IMO, IOI, and ICPC World Finals. Nvidia has ...
Harvard Business School

How Post-Training Reshapes LLMs: A Mechanistic View on Knowledge, Truthfulness, Refusal, and Confidence

Lakkaraju, Himabindu, Hongzhe Du, Weikai Li, Min Cai, Karim Saraipour, Zimin Zhang, Yizhou Sun, and Shichang Zhang. "How Post-Training Reshapes LLMs: A Mechanistic View on Knowledge, Truthfulness, ...