FCP: A Simple Training Trick to Make Autoregressive Video Generation Physically Plausible

Nanjing University and Tencent ARC Lab propose Future Chunk Prediction (FCP), an auxiliary training objective that lets autoregressive video diffusion models independently predict multiple future chunks, improving the physical plausibility of generated videos.

2026-10-22 ~ 2026-10-22 · 3 related posts

• Tencent ARC's Future Chunk Prediction Boosts Physical Plausibility in Video Diffusion — TencentARC · 2026-10-22
• Future Chunk Prediction: teaching autoregressive video models real physics — yshan2u · 2026-10-22
• Looking Ahead: Predicting Future Chunks Independently Makes Video Generation More Physical — yshan2u · 2026-10-22