Skip to content

๐ŸŽฌ Video Generation

๐Ÿ’ฌ EMNLP2026 ยท 1 paper notes

๐Ÿ“Œ Same area in other venues: ๐Ÿง  NeurIPS2026 (90) ยท ๐ŸŽž๏ธ ECCV2026 (176) ยท ๐Ÿ“ท CVPR2026 (188) ยท ๐Ÿ”ฌ ICLR2026 (99) ยท ๐Ÿ’ฌ ACL2026 (4) ยท ๐Ÿงช ICML2026 (32)

Helping Figures Tell their Story! Paper-Grounded Video Generation Explaining Complex Scientific Figures

To alleviate the high cognitive load of understanding dense scientific architecture diagrams, this paper establishes FigTalk, the first paper-grounded figure-to-video benchmark, and introduces MINARD, a three-team decoupled pipeline that coordinates narration generation, figure perception, and constrained visual grounding to synthesize faithful, temporally synchronized walkthrough videos.