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.
142-page report diving into the reasoning chains of R1. It spans 9 unique axes: safety, world modeling, faithfulness, long context, etc.
Comments
Paper: https://arxiv.org/abs/2504.07128
Repo: https://github.com/mcgill-NLP/thoughtology