DeepSeek-R1 Thoughtology: Let’s about LLM reasoning

142-page report diving into the reasoning chains of R1. It spans 9 unique axes: safety, world modeling, faithfulness, long context, etc.
Post image

Comments