Research map / Probabilistic and scientific modeling / Causal discovery
Dynamic Sparsity: Challenging Common Sparsity Assumptions for Learning World Models in Robotic Reinforcement Learning Benchmarks
Muthukumar Pandaram, Jakob Hollenstein, David Drexel, Samuele Tosatto, Antonio Rodriguez-Sanchez, Justus Piater
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 ground-truth Jacobian MAE loss (ablation) | MLP dynamics model · MLP world model on MuJoCo Playground | next-state prediction error | No clear effect |
| add ground-truth Jacobian MSE loss (ablation) | MLP dynamics model · MLP world model on MuJoCo Playground | next-state prediction error | No clear effect |
| add ground-truth Jacobian losses (ablation) | MLP dynamics model · MLP world model on MuJoCo Playground | induced Jacobian sparsity | Improved |
| add sparsity-aware error (SAE) loss on expected-zero Jacobian entries (ablation) | MLP dynamics model · MLP world model on MuJoCo Playground | induced sparsity and prediction error | Improved |
| add L1 penalty on predicted Jacobians (ablation) | MLP dynamics model · MLP world model on MuJoCo Playground | next-state prediction accuracy | Hurt |
| reduce MLP depth from 6 to 2 layers (ablation) | MLP dynamics model · MLP hyperparameter search 1 | mean test rank (lower better): 533.62 -> 245.32 | Improved |
| reduce dropout from 0.50 to 0.10 (ablation) | MLP dynamics model · MLP hyperparameter search 1 | mean test rank (lower better): 563.64 -> 237.48 | Improved |
| use learning rate 0.001 instead of 0.01 (ablation) | MLP dynamics model · MLP hyperparameter search 1 | mean test rank (lower better): 515.75 -> 338.43 | Improved |
| increase layer width from 32 to 512 (ablation) | MLP dynamics model · MLP hyperparameter search 1 | mean test rank (lower better): 420.80 -> 396.23 | Improved |
| use ELU instead of leaky-ReLU (ablation) | MLP dynamics model · MLP hyperparameter search 3 | mean test rank (lower better): 12.80 -> 8.05 | Improved |
| increase width from 512 to 1024 (ablation) | MLP dynamics model · MLP hyperparameter search 2 | mean test rank (lower better): 19.75 -> 21.25 | No clear effect |
Closest papers by meaning
- Reinforcement Learning Finetunes Small Subnetworks in Large Language Models (NeurIPS 2025 · similarity 0.88)
- Fine-Grained Causal Dynamics Learning with Quantization for Improving Robustness in Reinforcement Learning (ICML 2024 · similarity 0.87)
- Reinforcement Learning Under Latent Dynamics: Toward Statistical and Algorithmic Modularity (NeurIPS 2024 · similarity 0.87)
- Masked Diffusion Language Models are Strong and Steerable Text-Based World Models for Agentic RL (NeurIPS 2026 · similarity 0.87)
- Simplifying Latent Dynamics with Softly State-Invariant World Models (NeurIPS 2024 · similarity 0.87)
- Reward-Consistent Dynamics Models are Strongly Generalizable for Offline Reinforcement Learning (ICLR 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.
