🔓 Fully open-source ➡️
We’re releasing everything, including 4 staged models and 3 training datasets 🚀
🌐 Project page: http://ucsc-vlaa.github.io/VLM-CapCurriculum
📄 Paper: http://arxiv.org/abs/2605.20177
💻 Code: http://github.com/UCSC-VLAA/VLM-CapCurriculum
🤗 HF Collection: UCSC-VLAA/VLM-CapCurriculum
Built with amazing collaborators @ucsc @amazon @UWaterloo @VectorInst
If you’re working on VLM post-training, RLVR, or multimodal reasoning, we’d love to hear your thoughts! 👀
Again Kudos to my incredible PhD student, the leading author @JJwu41867797, as well as the whole team @HardyChen266091, @HaoqinT, Xianfeng Tang, @fredahshi, Hui Liu, Hanqing Lu, @cihangxie