Research map / Probabilistic and scientific modeling / Molecules and proteins
TrajEvo: Trajectory Prediction Heuristics Design via LLM-driven Evolution
Zhikai Zhao, Chuanbo Hua, Federico Berto, Kanghoon Lee, Zihan Ma, Jiachen Li, Jinkyoo Park
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 Cross-Generation Elite Sampling (ablation) | TrajEvo · ETH-UCY ETH split | minADE20 (m): 0.59 -> 0.47 | Improved |
| add Statistics Feedback Loop (ablation) | TrajEvo · ETH-UCY ETH split | minADE20 (m): 0.68 -> 0.59 (with CGES vs without both) | Improved |
| add SFL and CGES on easier splits (ZARA1) (ablation) | TrajEvo · ETH-UCY ZARA1 split | minADE20 (m): 0.36 -> 0.35 | Mixed |
| use TrajEvo heuristic instead of CVM-S | heuristic trajectory predictors · ETH-UCY avg | minADE20 (m): 0.47 -> 0.35 | Improved |
| use more complex heuristic (SocialForce, CTRV, ConstantAcc) than constant velocity | heuristic baselines · ETH-UCY avg | minADE20 (m): 0.47 (CVM-S) vs 1.06 (SocialForce) | Hurt |
| use TrajEvo heuristic instead of deep models on OOD SDD | trajectory prediction · SDD (OOD) | minADE20 (pixels): 14.55 (EigenTrajectory) -> 12.59 | Improved |
| use TrajEvo heuristic instead of MoFlow in-distribution | trajectory prediction · ETH-UCY avg | minADE20 (m): 0.20 -> 0.35 | Hurt |
| change generator LLM (DeepSeek-R1, V3, Qwen3-32B, GPT-4o, Claude 3.7) (ablation) | TrajEvo · ETH-UCY avg | minADE20 (m): 0.34 to 0.37 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- MoFlow — MoFlow: One-Step Flow Matching for Human Trajectory Forecasting via Implicit Maximum Likelihood Estimation based Distillation (CVPR 2025) · better on 1 of 3 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
- TrajAgent: An LLM-Agent Framework for Trajectory Modeling via Large-and-Small Model Collaboration (NeurIPS 2025 · similarity 0.89)
- Towards Predicting Any Human Trajectory In Context (NeurIPS 2025 · similarity 0.87)
- AutoTraces: Autoregressive Trajectory Forecasting via Multimodal Large Language Models (CVPR 2026 · similarity 0.86)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.86)
- Guided Trajectory Generation with Diffusion Models for Offline Model-based Optimization (NeurIPS 2024 · similarity 0.86)
- BAT: Behavior-Aware Human-Like Trajectory Prediction for Autonomous Driving (AAAI 2024 · 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.
