Research map / Video / Video generation
ActionParty: Multi-Subject Action Binding in Generative Video Games
Alexander Pondaven, Ziyi Wu, Igor Gilitschenski, Philip Torr, Sergey Tulyakov, Fabio Pizzati, Aliaksandr Siarohin
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 subject state tokens with attention masking and RoPE bias vs text-only action control | Wan2.1-1.3B video world model · Melting Pot (46 games) | Movement Accuracy: 0.158 -> 0.779 | Improved |
| Same, on visual fidelity | Wan2.1-1.3B video world model · Melting Pot (46 games) | FVD: 56.74 -> 17.16 | Improved |
| Same, on Interact effect accuracy | Wan2.1-1.3B video world model · Melting Pot (46 games) | Effect Accuracy (Interact): 0.326 -> 0.774 | Improved |
| Add cross-attention mask M_CA (ablation) | ActionParty · Coins | Movement Accuracy: 0.052 -> 0.872 | Improved |
| Add self-attention mask M_SA (ablation) | ActionParty · Coins | Movement Accuracy: 0.580 -> 0.872 | Improved |
| Add RoPE bias in self-attention linking subject tokens to video positions (ablation) | ActionParty · Coins | Movement Accuracy: 0.032 -> 0.872 | Improved |
| Use full-sequence M_CA instead of frame-wise M_CA (ablation) | ActionParty · Coins | Movement Accuracy: 0.052 -> 0.872 | Improved |
| Remove M_SA effect on subject preservation (ablation) | ActionParty · Coins | Subject Preservation: 1.00 -> 1.00 | No clear effect |
| Increase players from sparse to crowded | ActionParty · Melting Pot | Movement Accuracy: 0.781 -> 0.797 | No clear effect |
| Increase players from sparse to crowded | Text-Action · Melting Pot | Movement Accuracy: 0.196 -> 0.125 | Hurt |
Closest papers by meaning
- From Zero to Hero: Training-Free Custom Concept Spawning in World Models (NeurIPS 2026 · similarity 0.88)
- GameGen-X: Interactive Open-world Game Video Generation (ICLR 2025 · similarity 0.88)
- Genie: Generative Interactive Environments (ICML 2024 · similarity 0.88)
- Diffusion Models Are Real-Time Game Engines (ICLR 2025 · similarity 0.87)
- World Models as Group Actions (NeurIPS 2026 · similarity 0.87)
- Exploration-Driven Generative Interactive Environments (CVPR 2025 · similarity 0.86)
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.
