Research map / Video / Video segmentation and tracking
Long-Horizon Visual Planning with Goal-Conditioned Hierarchical Predictors
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 hierarchical GCP-tree instead of GCP-sequential | goal-conditioned video prediction · Human 3.6M | PSNR: 27.57 -> 28.34 | Helped |
| use hierarchical GCP-tree instead of GCP-sequential | goal-conditioned video prediction · 25-room Nav | PSNR: 11.61 -> 12.88 | Helped |
| use hierarchical GCP-tree instead of GCP-sequential | goal-conditioned video prediction · Pick&Place | PSNR: 34.45 -> 34.34 | Hurt |
| use hierarchical GCP-tree instead of GCP-sequential | goal-conditioned video prediction · 9-room Nav | PSNR: 12.91 -> 13.83 | Helped |
| add skip connections (ablation) | GCP-tree · Pick&Place | PSNR: $32.64$ -> $34.34$ | Helped |
| add an LSTM predictive module (ablation) | GCP-tree · Pick&Place | PSNR: $31.44$ -> $34.34$ | Helped |
| use hierarchical planning instead of GCP-flat planning | GCP-tree visual navigation planner · 25-room Nav | Success: 79% -> 82% | Helped |
| use hierarchical planning instead of GCP-flat planning | GCP-tree visual navigation planner · 25-room Nav | Cost: 181.02 -> 158.06 | Helped |
| use GCP-tree with hierarchical planning instead of GCP-sequential | visual navigation planner · 25-room Nav | Success: 14% -> 82% | Helped |
| use adaptive binding | GCP-tree video prediction · evaluated datasets | prediction quality | No clear effect |
| use GCP-tree instead of GCP-sequential | goal-conditioned prediction runtime · H3.6M sequences, length 1000 | Runtime (sec/training batch): $17.6$ -> $2.77$ | Helped |
| plan trajectories with GCP-tree instead of using the training-data trajectories | navigation trained on original data · 9-room Nav | Average trajectory length: $31.4$ -> $20.7$ | Helped |
Closest papers by meaning
- Hierarchical Foresight: Self-Supervised Learning of Long-Horizon Tasks via Visual Subgoal Generation (ICLR 2020 · similarity 0.91)
- Sparse Planning in Visual World Models via Cost Gradients (NeurIPS 2026 · similarity 0.89)
- Visual Planning: Let's Think Only with Images (ICLR 2026 · similarity 0.88)
- Efficient Robotic Policy Learning via Latent Space Backward Planning (ICML 2025 · similarity 0.88)
- PcLast: Discovering Plannable Continuous Latent States (ICML 2024 · similarity 0.88)
- Predictive but Not Plannable: RC-aux for Latent World Models (NeurIPS 2026 · similarity 0.88)
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.