[📂 GitHub] [📜 Paper]
This multimodal reasoning dataset extends the MM-Eureka dataset mainly by including reasoning answers, which are distilled from our finetuned Qwen2.5-VL-7B model. We provide two formats: llama_factory and verl for corresponding training frameworks.
Eureka-Distill is released under the Apache License 2.0. It is derived from MMK12/MM-Eureka, which is also released under Apache-2.0. Academic research use, including… See the full description on the dataset page:
https://huggingface.co/datasets/JierunChen/Eureka-Distill.
