Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Exploring Social Posterior Collapse in Variational Autoencoder for Interaction Modeling
Chen Tang, Wei Zhan, Masayoshi Tomizuka
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 |
|---|---|---|---|
| switch VAE to conditional prior (CVAE) only (ablation) | Social-CVAE · INTERACTION val HM | Agent Ratio AR (%): 8.68 -> 5.42 | Hurt |
| switch VAE to CVAE only (ablation) | Social-CVAE · INTERACTION val IS | Agent Ratio AR (%): 1.85 -> 0.18 | No clear effect |
| add auxiliary task to CVAE (full social-CVAE) (ablation) | Social-CVAE · INTERACTION val HM | Agent Ratio AR (%): 5.42 -> 15.5 | Improved |
| add auxiliary task to CVAE (ablation) | Social-CVAE · INTERACTION val HM K=6 | minFDE: 0.40 -> 0.32 | Improved |
| add auxiliary task to CVAE (ablation) | Social-CVAE · INTERACTION val IS K=6 | minFDE: 0.68 -> 0.70 | No clear effect |
| full social-CVAE vs instance that still collapses (Ours*) (ablation) | Social-CVAE · INTERPRET generalization track K=6 | ADE: 1.1615 -> 0.9205 | Improved |
| full social-CVAE vs Ours* (ablation) | Social-CVAE · INTERPRET regular track K=6 | ADE: 0.5157 -> 0.4665 | Improved |
| CVAE vs VAE (ablation) | Social-CVAE · INTERPRET generalization track K=6 | ADE: 1.3428 -> 1.4517 | Hurt |
| add auxiliary task to CVAE (ablation) | Social-CVAE · Argoverse val K=6 | minFDE: 2.09 -> 1.98 | Improved |
| add auxiliary task to CVAE (ablation) | Social-CVAE · Argoverse val | Agent Ratio AR (%): 0.40 -> 28.5 | Improved |
| switch VAE to CVAE (ablation) | Social-CVAE · ETH/UCY average | Agent Ratio AR (%): 34.7 -> 71.1 | Improved |
| add auxiliary task to CVAE (ablation) | Social-CVAE · ETH/UCY average | minADE: 0.31 -> 0.33 | No clear effect |
Closest papers by meaning
- Interaction Modeling with Multiplex Attention (NeurIPS 2022 · similarity 0.87)
- SocialCVAE: Predicting Pedestrian Trajectory via Interaction Conditioned Latents (AAAI 2024 · similarity 0.87)
- Social ODE: Multi-agent Trajectory Forecasting with Neural Ordinary Differential Equations (ECCV 2022 · similarity 0.87)
- Sim-to-Real Causal Transfer: A Metric Learning Approach to Causally-Aware Interaction Representations (CVPR 2025 · similarity 0.86)
- Beyond Vanilla Variational Autoencoders: Detecting Posterior Collapse in Conditional and Hierarchical Variational Autoencoders (ICLR 2024 · similarity 0.86)
- Poly-Autoregressive Prediction for Modeling Interactions (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.
