Research map / Reinforcement learning / Visual and representation RL
Look where you look! Saliency-guided Q-networks for generalization in visual Reinforcement Learning
David Bertoin, Adil Zouitine, Mehdi Zouitine, Emmanuel Rachelson
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 auxiliary self-supervised saliency prediction phase to SAC (ablation) | SAC · DMControl-GB video easy avg | episode return: 321 -> 789 | Improved |
| Add auxiliary self-supervised saliency prediction phase to SAC (ablation) | SAC · DMControl-GB video hard avg | episode return: 126 -> 375 | Improved |
| Add auxiliary self-supervised phase to SAC (ablation) | SAC · DMControl-GB train avg | episode return: 543 -> 940 | Improved |
| Add consistency regularization to critic loss of SAC (ablation) | SAC · DMControl-GB train avg | episode return: 543 -> 781 | Improved |
| Add consistency regularization to critic loss of SAC (ablation) | SAC · DMControl-GB video easy avg | episode return: 321 -> 647 | Improved |
| Add consistency regularization to critic loss of SAC (ablation) | SAC · DMControl-GB video hard avg | episode return: 126 -> 362 | Improved |
| Combine consistency loss with self-supervised phase (full SGQN) vs SAC+Self learning (ablation) | SAC · DMControl-GB video hard avg | episode return: 375 -> 747 | Improved |
| Combine both losses vs SAC+Self learning on video easy (ablation) | SAC · DMControl-GB video easy avg | episode return: 789 -> 906 | Improved |
| Add consistency loss on Cartpole training (ablation) | SAC · Cartpole train | episode return: 850 -> 839 | No clear effect |
| Lower mask threshold rho from 0.98 to 0.95 (ablation) | SGQN · Ball in cup video easy | episode return: 950 -> 905 | Hurt |
| Lower mask threshold rho from 0.98 to 0.95 (ablation) | SGQN · Cartpole* video hard | episode return: 569 -> 457 | Hurt |
| Lower mask threshold rho from 0.98 to 0.95 (ablation) | SGQN · Ball in cup video hard | episode return: 782 -> 789 | No clear effect |
Closest papers by meaning
- Learning Generalizable Agents via Saliency-guided Features Decorrelation (NeurIPS 2023 · similarity 0.89)
- A Simple Framework for Generalization in Visual RL under Dynamic Scene Perturbations (NeurIPS 2024 · similarity 0.88)
- Pre-Trained Image Encoder for Generalizable Visual Reinforcement Learning (NeurIPS 2022 · similarity 0.88)
- Focus On What Matters: Separated Models For Visual-Based RL Generalization (NeurIPS 2024 · similarity 0.87)
- What Effects the Generalization in Visual Reinforcement Learning: Policy Consistency with Truncated Return Prediction (AAAI 2024 · similarity 0.87)
- VIGOR: Zero-Shot Visual Generalization via Latent-Space Consistency in Model-Based Reinforcement Learning (NeurIPS 2026 · similarity 0.86)
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.
