DeepSeek R2 Model and SPCT Scaling Technique

DeepSeek AI published a research paper introducing SPCT, a novel technique to improve scalability of general reward models (GRMs) during inference for their next-gen R2 model.

One continuously updated timeline instead of dozens of separate articles. New developments are appended as the story evolves.

1. AnnouncementApr 11, 2025, 02:43 PM 76%

   DeepSeek unveils SPCT, a novel approach to scaling inference for its next-gen R2 model via general reward models (GRMs).

   DeepSeek AI published a research paper introducing SPCT, a novel technique to improve scalability of general reward models (GRMs) during inference for their next-gen R2 model.

   Read the full story →