AI 终于能边看边说，不再等你看完才开口
现在的 AI 视频对话是「先看完、再回答」:它得等你把整段视频播完，才开口。MOSS-VL 把「边看边说」当成头等能力——它生成回答的同时，眼睛还在看新画面，靠一种「门控交叉注意力」让视觉信息随时插进来，不用等。训练上，它用合成的对话语料教模型什么时候该说话、什么时候该闭嘴、什么时候该改口，再把实时相关的训练集中到最后一步，不破坏它原本的离线能力。结果：在 4 个流式评测里，它拿了 3 个第一，尤其在「主动提醒」这类考验模型自己判断时机的任务上，66.0 分对最强基线 37.5 分，近乎翻倍。而且它把视觉信息放在生成序列之外，首字延迟比同规模的 Qwen3-VL 快 2.8 到 5.1 倍。这不是你明天能装进手机的功能，但它是「AI 陪你视频聊天」从呆等变成真对话的关键一步。
📄 原文摘要(英文)
We present MOSS-VL, an open vision-language model family that treats real-time interaction -- perceiving while it speaks -- as a first-class capability. It is co-designed across the stack: the language decoder attends to vision only through gated cross-attention, so the model can naturally see incoming frames while generating; a synthesized interaction corpus supervises when to speak, when to stay silent, and when to revise; and a staged curriculum concentrates all real-time-specific training in one light final stage over a strong offline foundation. Offline, MOSS-VL-Instruct is competitive at comparable scale and leads temporal-reasoning video sets. Across four streaming benchmarks, MOSS-VL-Realtime posts the best average on three (second on the fourth) among open-source streaming models, sweeping the three subsets that squarely test proactive behavior -- 66.0 vs. 37.5 for the best baseline on OmniMMI Proactive Alerting. With 11.3B parameters but visual tokens outside the decoded sequence, MOSS-VL widens its time-to-first-token advantage over same-backbone Qwen3-VL-8B from 2.8x to 5.1x as visual context grows. We release all five checkpoints, the training curriculum, and the real-time inference code at https://github.com/OpenMOSS/MOSS-VL.