- Improve the performance of language and multimodal large models on 2D/3D generation and understanding tasks.
- Publish high-quality research papers and open-source core algorithms, code, and models to expand community impact.
- Solid background of language and multimodal models, with hands-on experience of pre-training or fine-tuning open-source models such as LLaMA or Qwen.
- Research experience or strong interest in one or more of the following areas: 2D image/graphic design, 3D scenes, or 3D CAD.
- Strong academic writing skills; prior publications in top-tier conferences (NeurIPS, ICML, ICLR, CVPR, ECCV, ICCV) are preferred.
- 提升语言大模型及多模态大模型在 2D/3D 生成与理解任务中的整体性能。
- 撰写高质量科研论文,并开源核心算法、代码与模型,扩大社区影响力。
- 熟练掌握语言大模型/多模态大模型原理,具备对 LLaMA、Qwen 等开源模型进行预训练或微调的实战经验。
- 对以下方向之一(或多项)具有研究经验或浓厚兴趣:2D图片/平面设计,3D场景,3D CAD。
- 具备扎实的学术写作能力;曾在 NeurIPS、ICML、ICLR、CVPR、ECCV、ICCV 等国际顶级会议发表论文者优先。