Current Video Large Language Models (Video LLMs) typically encode frames via a vision encoder and employ an autoregressive (AR) LLM for understanding and generation. However, this AR paradigm inevitably faces a dual efficiency bottleneck: strictly unidirectional attention compromises understanding efficiency by hindering global spatiotemporal aggregation, while serial decoding restricts generation efficiency. To address this, we propose VidLaDA, a Video LLM based on Diffusion Language Models (DLMs) that leverages bidirectional attention to unlock comprehensive spatiotemporal modeling and decode tokens in parallel. To further mitigate the computational overhead of diffusion decoding, we introduce MARS-Cache, an acceleration strategy that prunes redundancy by combining asynchronous visual cache refreshing with frame-wise chunk attention. Experiments show VidLaDA rivals state-of-the-art AR baselines (e.g., Qwen2.5-VL and LLaVA-Video) and outperforms DLM baselines, with MARS-Cache delivering over 12x speedup without compromising accuracy. Code and checkpoints are open-sourced at https://github.com/ziHoHe/VidLaDA.

翻译：当前视频大语言模型通常通过视觉编码器对视频帧进行编码，并采用自回归大语言模型进行理解与生成。然而，这种自回归范式不可避免地面临双重效率瓶颈：严格单向注意力机制因阻碍全局时空聚合而影响理解效率，同时串行解码限制了生成效率。为解决这一问题，我们提出VidLaDA，一种基于扩散语言模型的视频大语言模型，其利用双向注意力机制实现全面的时空建模并支持并行令牌解码。为降低扩散解码的计算开销，我们进一步提出MARS-Cache加速策略，该方法通过异步视觉缓存刷新与帧级分块注意力相结合来剪除冗余计算。实验表明，VidLaDA在性能上可与先进的自回归基线模型（如Qwen2.5-VL和LLaVA-Video）相媲美，并显著超越现有扩散语言模型基线，同时MARS-Cache在保持精度无损的情况下实现了超过12倍的加速效果。代码与模型权重已开源发布于https://github.com/ziHoHe/VidLaDA。