Research map / Detection and segmentation / Open-set and few-shot detection
A Critical View of Vision-Based Long-Term Dynamics Prediction Under Environment Misalignment
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 segmentation mask (GT-Mask) instead of raw RGB as input | RPCIN trained on SimB-Border · Cross-Domain SimB-Border -> BlenB-Border | P2 squared l2 (x1000): 23.564 (RGB-BN) -> 9.358 (GT-Mask) | Helped |
| Use segmentation mask (GT-Mask) instead of raw RGB as input | RPCIN trained on SimB-Split · Cross-Domain SimB-Split -> BlenB-Split | P2 squared l2 (x1000): 27.511 (RGB-BN) -> 7.431 (GT-Mask) | Helped |
| Use self-supervised k-means mask (Self-Mask) instead of raw RGB | RPCIN trained on SimB-Split · Cross-Domain SimB-Split -> BlenB-Split | P2 squared l2 (x1000): 27.511 -> 8.017 | Helped |
| Use self-supervised k-means mask (Self-Mask) instead of raw RGB | RPCIN trained on SimB-Border · Cross-Domain SimB-Border -> BlenB-Border | P2 squared l2 (x1000): 23.564 -> 9.614 | Helped |
| Use GT or self-supervised mask instead of raw RGB | RPCIN trained on BlenB-Border evaluated on SimB-Border · Cross-Domain BlenB-Border -> SimB-Border | P1 squared l2 (x1000): 261.768 (RGB-BN) -> 1.091 (GT-Mask), 1.136 (Self-Mask) | Helped |
| Use GT or self-supervised mask instead of raw RGB | RPCIN trained on BlenB-Split evaluated on SimB-Split · Cross-Domain BlenB-Split -> SimB-Split | P1 squared l2 (x1000): 171.750 (RGB-BN) -> 0.916 (GT-Mask), 0.944 (Self-Mask) | Helped |
| Replace BatchNorm with GroupNorm/LayerNorm on raw RGB input | RPCIN cross-domain transfer · Cross-Domain SimB-Border -> BlenB-Border | P2 squared l2 (x1000): 23.564 (BN) -> 11.507 (GN) | Helped |
| Use self-supervised mask in place of GT mask when training on source mask (ablation) | Mask generality: GT-Mask train -> Self-Mask test, Sim->Blen · SimB-Split -> BlenB-Split | P2 squared l2 (x1000): 7.431 -> 8.576 | Hurt |
| Increase alignment loss weight on predicted state feature (ablation) | RPCIN with GT-Mask, SimB-Split -> SimB-Border · Cross-Context | performance gap | No clear effect |
| Increase RoI pooling output size (ablation) | RPCIN with GT-Mask · Cross-Context | performance gap | No clear effect |
| Train and test in Split context vs Border context (context change alone) | RPCIN with RGB-BN · Cross-Context SimB-Border -> SimB-Split | P2 squared l2 (x1000): 9.568 aligned -> 39.773 cross | Hurt |
Closest papers by meaning
- StateSpaceDiffuser: Bringing Long Context to Diffusion World Models (NeurIPS 2025 · similarity 0.84)
- DynaTokens: Teaching Dynamics to Camera-Controlled Video Models at Test Time (NeurIPS 2026 · similarity 0.84)
- Video World Models with Long-term Spatial Memory (NeurIPS 2025 · similarity 0.84)
- PlaySlot: Learning Inverse Latent Dynamics for Controllable Object-Centric Video Prediction and Planning (ICML 2025 · similarity 0.84)
- Learning Visual Feature-Based World Models via Residual Latent Action (NeurIPS 2026 · similarity 0.84)
- Trimming the Long-Tail of Visual World Modeling Evaluation (NeurIPS 2026 · similarity 0.84)
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.