WorldDiT: A Unified Diffusion Architecture for World and Action Modeling
- 类型:arxiv
- 标识:2607.23909
- 链接:https://arxiv.org/abs/2607.23909
- 主分类:multimodal
- 形态:method
- 被引:0
- 被引来源:Semantic Scholar + OpenAlex
- S2被引:0
- OpenAlex被引:0
- 影响力被引:0
- TLDR:WorldDiT, a unified diffusion transformer architecture that couples action generation with visual world modeling and achieves strong performance without a large pretrained VLM action backbone, lies on the reported Pareto frontier for total model parameters and mean success among methods reporting all four suites.
- OpenAlex ID:W7171512290
- OpenAlex DOI:10.48550/arxiv.2607.23909
- DOI:10.48550/arxiv.2607.23909
- DOI来源:OpenAlex
- 开放获取:green
- 开放获取链接:https://doi.org/10.48550/arxiv.2607.23909
- OpenAlex更新:2026-08-25
- 副分类:engineering
- 待LLM分类:否
- 标题中文:WorldDiT:面向世界建模与动作建模的统一扩散架构
- TLDR中文:WorldDiT 是一种统一的 diffusion Transformer 架构,将动作生成与视觉世界建模耦合,无需大型预训练 VLM 动作主干即取得强性能,在报告全部四个 suite 的方法中,其总模型参数量与平均成功率处于已报告的 Pareto 前沿上。
- 来源文件:
- /inbox/tom/_candidates/2026-07-29-agent-rag-longcontext-candidates.json
- [S2 enrich]
- [OpenAlex backfill]