Research map / Video / Video generation
Anchoring and Rescaling Attention for Semantically Coherent Inbetweening
Tae Eun Choi, Sumin Shim, Junhyeok Kim, Seong Jae Hwang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add KAB (keyframe-anchored attention bias) alone (ablation) | Wan2.1 FLF2V · TGI-Bench (81 frames) | PSNR: 17.63 -> 17.62 | No clear effect |
| add KAB alone (ablation) | Wan2.1 FLF2V · TGI-Bench (81 frames) | VBench: 9.904 -> 9.896 | No clear effect |
| add ReTRo (rescaled temporal RoPE) alone (ablation) | Wan2.1 FLF2V · TGI-Bench (81 frames) | PSNR: 17.63 -> 18.18 | Improved |
| add ReTRo alone (ablation) | Wan2.1 FLF2V · TGI-Bench (81 frames) | FID: 82.90 -> 77.88 | Improved |
| add ReTRo alone (ablation) | Wan2.1 FLF2V · TGI-Bench (81 frames) | VBench: 9.904 -> 9.749 | Hurt |
| add KAB on top of ReTRo (ablation) | Wan2.1 FLF2V + ReTRo · TGI-Bench (81 frames) | VBench: 9.749 -> 10.022 | Improved |
| add KAB on top of ReTRo (ablation) | Wan2.1 FLF2V + ReTRo · TGI-Bench (81 frames) | PSNR: 18.18 -> 18.17 | No clear effect |
| add KAB + ReTRo (full method) | Wan2.1 FLF2V · TGI-Bench (81 frames) | FID: 82.90 -> 77.59 | Improved |
| add KAB + ReTRo (full method) | Wan2.1 FLF2V · TGI-Bench (81 frames) | Human pace stability: 3.65 -> 4.34 | Improved |
| add KAB + ReTRo (full method) | Wan2.1 FLF2V · TGI-Bench (81 frames) | Human semantic fidelity: 3.69 -> 4.27 | Improved |
| add KAB + ReTRo (full method) | Wan2.1 FLF2V · TGI-Bench (65 frames) | FVD: 0.3406 -> 0.2820 | Improved |
Closest papers by meaning
- Generative Inbetweening through Frame-wise Conditions-Driven Video Generation (CVPR 2025 · similarity 0.88)
- Leveraging Unpaired Data for Vision-Language Generative Models via Cycle Consistency (ICLR 2024 · similarity 0.86)
- Towards Long-window Anchoring in Vision-Language Model Distillation (AAAI 2026 · similarity 0.86)
- Generative Inbetweening: Adapting Image-to-Video Models for Keyframe Interpolation (ICLR 2025 · similarity 0.85)
- Inference-Time Text-to-Video Alignment with Diffusion Latent Beam Search (NeurIPS 2025 · similarity 0.85)
- Enhanced Fine-Grained Motion Diffusion for Text-Driven Human Motion Synthesis (AAAI 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
