Light-R1
Collection
Curriculum SFT, DPO and RL for Long COT from Scratch and Beyond
•
7 items
•
Updated
•
9
Quantized models' evaluation results:
Models | AIME24 | AIME25 |
---|---|---|
Light-R1-14B-DS | 74.0 | 60.2 |
Light-R1-14B-DS-Q4_0.gguf (int4) | 70.1 | 54.9 |
Light-R1-14B-DS-Q8_0.gguf (int8) | 71.9 | 59.4 |
Light-R1-14B-DS-Q4-KM.gguf (q4-k-m) | 70 | 61.3 |
Base model
deepseek-ai/DeepSeek-R1-Distill-Qwen-14B