Research map / Reinforcement learning / Visual and representation RL
An Investigation into Pre-Training Object-Centric Representations for Reinforcement Learning
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 |
|---|---|---|---|
| add object-centric pre-training | image-based RL agents · Object Comparison and Property Comparison | success rate | Helped |
| add object-centric pre-training | image-based RL agents · Object Goal | success rate | No clear effect |
| add object-centric pre-training with non-SLATE OCR models | image-based RL agents · Object Interaction | success rate | Hurt |
| add object-centric pre-training | image-based RL agents · Object Comparison and Property Comparison | sample efficiency | Helped |
| add object-centric pre-training | image-based RL agents · Object Interaction | sample efficiency | Hurt |
| add object-centric pre-training | image-based RL agents · unseen object colors | success rate | Mixed |
| replace Transformer pooling with MLP pooling (ablation) | SLATE · Object Goal | success rate: 0.985 -> 0.9 | Hurt |
| replace Transformer pooling with MLP pooling (ablation) | SLATE · Object Interaction | success rate: 0.787 -> 0.03 | Hurt |
| replace Transformer pooling with MLP pooling (ablation) | SLATE · Object Comparison | success rate: 0.979 -> 0.238 | Hurt |
| replace Transformer pooling with MLP pooling (ablation) | SLATE · Property Comparison | success rate: 0.98 -> 0.229 | Hurt |
| replace Transformer pooling with Deep Sets pooling (ablation) | SLATE · Object Interaction | success rate: 0.787 -> 0.103 | Hurt |
| replace Transformer pooling with Relational Network pooling (ablation) | SLATE · Property Comparison | success rate: 0.980 -> 0.150 | Hurt |
Closest papers by meaning
- The Surprising Ineffectiveness of Pre-Trained Visual Representations for Model-Based Reinforcement Learning (NeurIPS 2024 · similarity 0.90)
- Investigating Pre-Training Objectives for Generalization in Vision-Based Reinforcement Learning (ICML 2024 · similarity 0.88)
- Efficient Reinforcement Learning Through Adaptively Pretrained Visual Encoder (AAAI 2025 · similarity 0.88)
- Entity-Centric Reinforcement Learning for Object Manipulation from Pixels (ICLR 2024 · similarity 0.88)
- When Should We Prefer State-to-Visual DAgger over Visual Reinforcement Learning? (AAAI 2025 · similarity 0.87)
- Focus On What Matters: Separated Models For Visual-Based RL Generalization (NeurIPS 2024 · similarity 0.87)
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.