Research map / Reinforcement learning / Visual and representation RL
GPLAC: Generalizing Vision-Based Robotic Skills Using Weakly Labeled Images
Avi Singh, Larry Yang, Sergey Levine
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 attention architecture to classifier-augmented CNN (ablation) | GPLAC · Pusher (simulation, unseen envs) | success rate (%): 2.25 -> 79.75 | Improved |
| add attention architecture to classifier-augmented CNN (ablation) | GPLAC · Striker (simulation, unseen envs) | success rate (%): 21.50 -> 59.25 | Improved |
| add weakly labeled data via binary classifier (ablation) | ATT-1 · Pusher (simulation, unseen envs) | success rate (%): 37.50 -> 79.75 | Improved |
| add weakly labeled data via binary classifier (ablation) | ATT-1 · Striker (simulation, unseen envs) | success rate (%): 35.50 -> 59.25 | Improved |
| add weakly labeled data via binary classifier (ablation) | ATT-1 · Real robot mug pushing | success rate (%): 33.22 -> 56.36 | Improved |
| add domain adaptation on top of GPLAC | GPLAC · Pusher (simulation, unseen envs) | success rate (%): 79.75 -> 66.50 | Hurt |
| add domain adaptation on top of GPLAC | GPLAC · Striker (simulation, unseen envs) | success rate (%): 59.25 -> 61.00 | Improved |
| use binary classifier instead of domain adaptation for weak data (ablation) | ATT-DA · Pusher (simulation, unseen envs) | success rate (%): 69.50 -> 79.75 | Improved |
| use binary classifier instead of domain adaptation for weak data (ablation) | ATT-DA · Striker (simulation, unseen envs) | success rate (%): 40.50 -> 59.25 | Improved |
| reduce weakly labeled images from 80000 to 800 (ablation) | GPLAC · Pusher (simulation, unseen envs) | success rate (%): 79.75 -> 78.50 | No clear effect |
| reduce weakly labeled images from 800 to 80 (ablation) | GPLAC · Pusher (simulation, unseen envs) | success rate (%): 78.50 -> 56.00 | Hurt |
| reduce weakly labeled images from 800 to 80 (ablation) | GPLAC · Striker (simulation, unseen envs) | success rate (%): 53.00 -> 33.50 | Hurt |
Closest papers by meaning
- Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers (NeurIPS 2024 · similarity 0.88)
- Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulation (ICLR 2024 · similarity 0.88)
- Exploring the Limits of Vision-Language-Action Manipulation in Cross-task Generalization (NeurIPS 2025 · similarity 0.88)
- STRAP: Robot Sub-Trajectory Retrieval for Augmented Policy Learning (ICLR 2025 · similarity 0.88)
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · similarity 0.88)
- 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.
