Research map / Video / Video generation
Instilling an Active Mind in Avatars via Cognitive Simulation
Jianwen Jiang, Weihong Zeng, Zerong Zheng, Jiaqi Yang, Chao Liang, Wang Liao, Han Liang, Weifeng Chen, XING WANG, Yuan Zhang, Mingyuan Gao
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 agentic MLLM reasoning (System 2) (ablation) | OmniHuman-1.5 · Custom single-subject set (150 cases) | HKV: 122.376 -> 168.912 | Improved |
| add agentic reasoning (ablation) | OmniHuman-1.5 · Pairwise user study | motion unnaturalness: 0.58 -> 0.37 | Improved |
| add agentic reasoning (ablation) | OmniHuman-1.5 · Custom single-subject set (150 cases) | IQA / Sync-C: 4.784 -> 4.790 IQA | No clear effect |
| use symmetric audio branch with shared attention instead of cross-attention (ablation) | OmniHuman-1.5 · Custom single-subject set (150 cases) | Sync-C: 3.263 -> 4.087 | Improved |
| add MM-branch warm-up (ablation) | OmniHuman-1.5 · Custom single-subject set (150 cases) | HKV: 164.080 -> 168.912 | Improved |
| use pseudo last frame instead of reference image conditioning (ablation) | OmniHuman-1.5 · Custom single-subject set (150 cases) | HKV: 160.889 -> 168.912 | Improved |
| remove both reference and pseudo frame (ablation) | OmniHuman-1.5 · Custom single-subject set (150 cases) | IQA: 4.790 -> 4.682 | Hurt |
| add reasoning to multi-person model (ablation) | OmniHuman-1.5 · Multi-person test set (57 cases) | HKV: 138.43 -> 158.36 | Improved |
| inject MLLM reasoning latents into synthesis (ablation) | OmniHuman-1.5 · Single-subject set | motion dynamics | Mixed |
| use our conditioning instead of OmniHuman-1 conditioning (ablation) | OmniHuman-1.5 · Pairwise user study | GSB: -0.23 -> +0.23 | Improved |
| compare with OmniHuman-1 on portrait | OmniHuman-1.5 · CelebV-HQ (portrait) | Sync-C: 5.199 -> 5.053 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- OmniHuman-1 — OmniHuman-1: Rethinking the Scaling-Up of One-Stage Conditioned Human Animation Models (ICCV 2025) · better on 7 of 12 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- InstructAvatar: Text-Guided Emotion and Motion Control for Avatar Generation (AAAI 2025 · similarity 0.90)
- VLOGGER: Multimodal Diffusion for Embodied Avatar Synthesis (CVPR 2025 · similarity 0.89)
- DAWN: Dynamic Frame Avatar with Non-autoregressive Diffusion Framework for Talking head Video Generation (ICLR 2025 · similarity 0.88)
- EMOPortraits: Emotion-enhanced Multimodal One-shot Head Avatars (CVPR 2024 · similarity 0.87)
- AvatarGO: Zero-shot 4D Human-Object Interaction Generation and Animation (ICLR 2025 · similarity 0.87)
- Make-Your-Anchor: A Diffusion-based 2D Avatar Generation Framework (CVPR 2024 · similarity 0.87)
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.
