Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Social NCE: Contrastive Learning of Socially-Aware Motion Representations
Yuejiang Liu, Qi Yan, Alexandre Alahi
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 Social-NCE loss | Social-STGCNN · ETH/UCY average | collision rate (%): 4.70 -> 2.96 | Improved |
| add Social-NCE loss | Trajectron++ · ETH/UCY average | collision rate (%): 1.37 -> 0.93 | Improved |
| add Social-NCE loss | Social-STGCNN · ETH/UCY average | top-20 FDE: 0.765 -> 0.756 | No clear effect |
| add Social-NCE loss | Trajectron++ · ETH/UCY average | top-20 FDE: 0.403 -> 0.398 | No clear effect |
| add Social-NCE loss | Social-LSTM · Trajnet++ overall | collision rate (%): 6.44 -> 5.31 | Improved |
| add Social-NCE loss | Directional-LSTM · Trajnet++ overall | collision rate (%): 5.49 -> 4.59 | Improved |
| add Social-NCE loss with social negatives | behavioral cloning · crowd navigation, 5k demos | collision rate (%): 11.11 -> 3.40 | Improved |
| replace social negatives with random negative sampling (ablation) | Social-NCE · crowd navigation, 5k demos | collision rate (%): 3.40 -> 18.60 (vanilla 11.11) | Hurt |
| add Social-NCE loss | Rainbow DQN offline · crowd navigation, 10% of data | normalized score (%): 14.4% -> 69.0% | Improved |
| add random-negative contrastive loss | Rainbow DQN offline · crowd navigation, 10% of data | normalized score (%): 14.4% -> 7.0% | Hurt |
| add Social-NCE loss | Rainbow DQN online · crowd navigation | episodes to reach reward 0.6: more than 4000 -> less than 2000 | Improved |
| use multiple sampling horizons 1-4 instead of single horizon 1 (ablation) | Social-NCE · behavioral cloning | collision rate (%): 11.24 -> 3.40 | Improved |
Closest papers by meaning
- SocialCVAE: Predicting Pedestrian Trajectory via Interaction Conditioned Latents (AAAI 2024 · similarity 0.88)
- Following the Human Thread in Social Navigation (ICLR 2025 · similarity 0.87)
- Social ODE: Multi-agent Trajectory Forecasting with Neural Ordinary Differential Equations (ECCV 2022 · similarity 0.87)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.86)
- Sim-to-Real Causal Transfer: A Metric Learning Approach to Causally-Aware Interaction Representations (CVPR 2025 · similarity 0.86)
- SocialCircle: Learning the Angle-based Social Interaction Representation for Pedestrian Trajectory Prediction (CVPR 2024 · similarity 0.85)
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.
