Research map / Video / Video generation
AnimeGamer: Infinite Anime Life Simulation with Next Game State Prediction
Junhao Cheng, Yuying Ge, Yixiao Ge, Jing Liao, Ying Shan
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 |
|---|---|---|---|
| use first frame instead of random frame for visual feature f_v (ablation) | AnimeGamer encoder · Ablation on tokenizer/de-tokenizer (Qiqi film, w/o MLLM) | ACC-F: 0.4744 -> 0.7293 | Improved |
| use MLP instead of single Linear layer in E_a (ablation) | AnimeGamer encoder · Ablation on tokenizer/de-tokenizer (Qiqi film, w/o MLLM) | ACC-F: 0.3649 -> 0.7293 | Improved |
| combine f_v and f_md by concatenation instead of element-wise addition (ablation) | AnimeGamer encoder · Ablation on tokenizer/de-tokenizer (Qiqi film, w/o MLLM) | CLIP-I: 0.7684 -> 0.8672 | Improved |
| combine f_v and f_md by concatenation instead of cross-attention (ablation) | AnimeGamer encoder · Ablation on tokenizer/de-tokenizer (Qiqi film, w/o MLLM) | ACC-F: 0.3284 -> 0.7293 | Improved |
| add warm-up training phase for E_a (ablation) | AnimeGamer encoder · Ablation on tokenizer/de-tokenizer (Qiqi film, w/o MLLM) | ACC-F: 0.7028 -> 0.7293 | Improved |
| add motion scope f_ms to the decoder D_a (ablation) | AnimeGamer decoder · Ablation on tokenizer/de-tokenizer (Qiqi film, w/o MLLM) | ACC-F: 0.1824 -> 0.7293 | Improved |
| add motion scope f_ms to the decoder D_a (ablation) | AnimeGamer decoder · Ablation on tokenizer/de-tokenizer (Qiqi film, w/o MLLM) | MAE-F: 1.2189 -> 0.4029 | Improved |
| add cosine similarity loss for MLLM training (ablation) | AnimeGamer · Ablation on next game state prediction/decoder adaptation (Qiqi film, with MLLM) | ACC-F: 0.6722 -> 0.6649 | No clear effect |
| add cosine similarity loss for MLLM training (ablation) | AnimeGamer · Ablation on next game state prediction/decoder adaptation (Qiqi film, with MLLM) | CLIP-I: 0.7856 -> 0.7628 | Hurt |
| add decoder adaptation training phase (ablation) | AnimeGamer · Ablation on next game state prediction/decoder adaptation (Qiqi film, with MLLM) | CLIP-I: 0.6831 -> 0.7856 | Improved |
| add decoder adaptation training phase (ablation) | AnimeGamer · Ablation on next game state prediction/decoder adaptation (Qiqi film, with MLLM) | ACC-F: 0.3649 -> 0.6722 | Improved |
| use action-aware multimodal MLLM game state prediction instead of fine-tuned CogVideoX (GC) | AnimeGamer vs GC · Anime life simulation benchmark (automatic) | ACC-F: 0.4249 -> 0.6744 | Improved |
Closest papers by meaning
- Unbounded: A Generative Infinite Game of Character Life Simulation (ICLR 2025 · similarity 0.88)
- GameGen-X: Interactive Open-world Game Video Generation (ICLR 2025 · similarity 0.86)
- From Zero to Hero: Training-Free Custom Concept Spawning in World Models (NeurIPS 2026 · similarity 0.84)
- PhysAnimator: Physics-Guided Generative Cartoon Animation (CVPR 2025 · similarity 0.83)
- APISR: Anime Production Inspired Real-World Anime Super-Resolution (CVPR 2024 · similarity 0.83)
- Animate-X: Universal Character Image Animation with Enhanced Motion Representation (ICLR 2025 · similarity 0.83)
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.
