Research map / Detection and segmentation / Open-set and few-shot detection
Object-Centric World Models from Few-Shot Annotations for Sample-Efficient Reinforcement Learning
Weipu Zhang, Adam Jelley, Trevor McInroe, Amos Storkey, Gang Wang
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 Cutie object vectors to STORM | STORM · Atari 100k (26 games) | HNS mean: 114.2% -> 134.8% | Improved |
| add Cutie object vectors to STORM | STORM · Atari 100k (26 games) | HNS median: 42.5% -> 43.8% | Improved |
| add Cutie object vectors to DreamerV3 | DreamerV3 · Atari 100k (26 games) | HNS mean: 106.6% -> 119.4% | Improved |
| add Cutie object vectors to DreamerV3 | DreamerV3 · Atari 100k (26 games) | HNS median: 33.2% -> 42.6% | Improved |
| use SAM2 instead of Cutie as object extractor in OC-STORM (ablation) | OC-STORM · Atari 100k (26 games) | HNS mean: 134.8% -> 124.6% | Hurt |
| use mask-based FOCUS representation instead of vector objects (ablation) | OC-STORM · Atari 100k (26 games) | HNS mean: 134.8% -> 107.2% | Hurt |
| add Cutie object vectors to STORM | STORM · Atari 100k object-detectable games | HNS: 147.7% -> 186.2% | Improved |
| add Cutie object vectors to STORM | STORM · Atari 100k other games | HNS: 80.8% -> 83.4% | No clear effect |
| add Cutie object vectors to STORM | STORM · Atari BattleZone | game score: 5830 -> 4590 | Hurt |
| add Cutie object vectors to STORM | STORM · Atari Hero | game score: 11434 -> 6441 | Hurt |
| add object vectors to STORM | STORM · Hollow Knight boss fights | training episode return | Improved |
| increase detection accuracy (lower random feature zeroing probability) (ablation) | OC-STORM object module only · Atari Boxing and Pong | training episode return | Improved |
Closest papers by meaning
- Learning Interactive World Model for Object-Centric Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Dyn-O: Building Structured World Models with Object-Centric Representations (NeurIPS 2025 · similarity 0.88)
- Efficient Exploration and Discriminative World Model Learning with an Object-Centric Abstraction (ICLR 2025 · similarity 0.88)
- Scaling Offline Model-Based RL via Jointly-Optimized World-Action Model Pretraining (ICLR 2025 · similarity 0.86)
- The Surprising Ineffectiveness of Pre-Trained Visual Representations for Model-Based Reinforcement Learning (NeurIPS 2024 · similarity 0.86)
- Zero-shot World Models via Search in Memory (NeurIPS 2025 · 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.
