Research map / Vision-language and multimodal / Robot action and navigation
Visual Grounding for Object-Level Generalization in Reinforcement Learning
Haobin Jiang, Zongqing Lu
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 |
|---|---|---|---|
| replace MineCLIP reward with object-grounded focal reward | PPO agent · hunt a cow | success rate (%): 3.8±4.8 -> 71.3±9.7 | Improved |
| add denoising (negative-word zeroing and 0.2 threshold) to confidence map (ablation) | Focal reward · hunt a cow | success rate (%): 3.8±4.8 -> 71.3±9.7 | Improved |
| use focal reward level instead of its step-to-step difference (ablation) | Focal reward · hunt a cow | success rate (%): 3.5±3.0 -> 71.3±9.7 | Improved |
| add Gaussian kernel centered on view to focal reward (ablation) | Focal reward · hunt a cow, two target animals | success rate (%): 14.5±21.8 -> 67.0±15.0 | Improved |
| condition policy on confidence map instead of language embedding | COPL · Hunting domain, unseen targets (avg) | success rate (%): 11.5±7.6 -> 48.1±6.6 | Improved |
| condition policy on confidence map instead of language embedding | COPL · Harvest domain, unseen targets (avg) | success rate (%): 16.7±4.3 -> 40.8±8.3 | Improved |
| condition policy on confidence map instead of language embedding | COPL · Harvest domain, training tasks (avg) | success rate (%): 73.2±5.2 -> 75.6±3.2 | No clear effect |
| condition policy on confidence map instead of language embedding | COPL · Hunting domain, training tasks (avg) | success rate (%): 16.6±13.3 -> 50.0±8.9 | Improved |
| COPL versus STEVE-1 on unseen spider target | COPL · hunt a spider (unseen) | success rate (%): 54.5±12.7 vs 73.0 | Hurt |
Closest papers by meaning
- What Can RL Bring to VLA Generalization? An Empirical Study (NeurIPS 2025 · similarity 0.90)
- Code as Reward: Empowering Reinforcement Learning with VLMs (ICML 2024 · similarity 0.90)
- Ground-Compose-Reinforce: Grounding Language in Agentic Behaviours using Limited Data (NeurIPS 2025 · similarity 0.89)
- Grounded Reinforcement Learning for Visual Reasoning (NeurIPS 2025 · similarity 0.89)
- GoalLadder: Incremental Goal Discovery with Vision-Language Models (NeurIPS 2025 · similarity 0.89)
- Vision-Language Models are Zero-Shot Reward Models for Reinforcement Learning (ICLR 2024 · similarity 0.89)
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.
