A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Shaltiel Shmidman, A. Fredman, Oleg Sudakov, et al.
This paper compares the accuracy and inference efficiency of medium-sized LLMs post-trained on reasoning traces from DeepSeek-R1 versus gpt-oss for math problems.
Zhengren Wang, Jiayang Yu, Dongsheng Ma, et al.
RARE decouples knowledge storage from reasoning by externalizing domain knowledge to retrievable sources and internalizing reasoning patterns, enabling lightweight models to surpass GPT-4 and DeepSeek-R1 by ~20% accuracy.
Mo Li, Songyang Zhang, Taolin Zhang, et al.
NeedleBench is a synthetic framework for evaluating LLMs' retrieval and reasoning in long contexts, revealing that reasoning models like Deepseek-R1 and o3 struggle with continuous retrieval in information-dense scenarios due to 'under-thinking'.
Kaiyan Zhang, Yuxin Zuo, Bingxiang He, et al.
A comprehensive survey of reinforcement learning for large reasoning models, covering foundational components, core problems, training resources, and downstream applications since DeepSeek-R1.
Unknown
A 14B parameter reasoning language model derived from Deepseek-R1-Distilled-Qwen-14B, trained with SFT and model merging, fully open-sourced.
Daya Guo, Dejian Yang, Haowei Zhang, et al.
DeepSeek-R1 shows that pure reinforcement learning can incentivize reasoning in LLMs without human-annotated trajectories, achieving superior performance on verifiable tasks.