Skip to content
Hot storiesContinuously updated

商汤与港科大等联合提出多轮图像编辑方法MT Forcing

1 reports1 reporting sourcesUpdated 23 hours ago

Start here

AI overview

2026年11月3日,商汤研究院联合香港科技大学 MMLab 等机构提出多轮图像编辑训练方法 MT Forcing。该方法主要针对自回归扩散模型在连续编辑过程中容易出现的画质退化与内容漂移等问题。技术方案上,MT Forcing 融合了 Single-Turn DMD 教师分布匹配与 Teacher-Forcing Guidance 监督机制,以此有效缓解暴露偏差,并在多轮编辑中更好地保留跨轮细节。

AI-generated from reports · Updated 23 hours ago

Report timeline

Follow the reports to see different sides of the story.

Nov 3
  1. AITNT · AI头条
    商汤与港科大 MMLab 等联合提出 MT Forcing,改善多轮图像编辑画质退化与内容漂移

    商汤研究院联合港科大 MMLab 等机构提出多轮图像编辑训练方法 MT Forcing,以解决自回归扩散模型在连续编辑中的画质退化与内容漂移。该方案融合 Single-Turn DMD 教师分布匹配与 Teacher-Forcing Guidance 监督,有效缓解暴露偏差并保留跨轮细节。

Story heat trend

Current heat 5·Comparable peak 10(Nov 3 17:00)·Comparable change, last 24 hours–

02.557.510Nov3Nov4Nov4Nov4

Trends compare only the same parties observed continuously, so the scope can be narrower than current heat counts. Move the pointer or tap the chart to see hourly heat; use the left and right arrow keys on a keyboard.