Research map / Multi-agent systems and games / Human-AI and goal-conditioned
LAGEA: Language Guided Embodied Agents for Robotic Manipulation
Abdul Monaf Chowdhury, Akm Moshiur Rahman Mazumder, Safaeid Hossain Arib, Rabeya Akter
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 structured schema-constrained feedback (ablation) | LaGEA · MT10 tasks | average success rate (%): 78.89 -> 98.89 | Improved |
| add LaGEA keyframe selection (vs uniform) (ablation) | LaGEA · MT10 five observable tasks | average success rate (%): 67.33 -> 80 | Improved |
| add LaGEA keyframe selection (vs random) (ablation) | LaGEA · MT10 five observable tasks | average success rate (%): 68 -> 80 | Improved |
| add learned feedback-vision alignment (ablation) | LaGEA · button-press-topdown-v2 observable | success rate (%): 20 -> 93.33 | Improved |
| add adaptive failure-aware rho (ablation) | LaGEA · button-press-topdown-v2 observable | success rate (%): 13.33 -> 93.33 | Improved |
| add delta (potential-based) rewards (ablation) | LaGEA · button-press-topdown-v2 observable | success rate (%): 30 -> 93.33 | Improved |
| add goal and feedback delta terms to adaptive rho (ablation) | LaGEA · drawer-open-v2 observable | success rate (%): 0 -> 100 | Improved |
| replace weaker VLM with Qwen-2.5-VL-3B (ablation) | LaGEA · MT10 five observable tasks | average success rate (%): 56 -> 80 | Improved |
| replace LIV text encoder with GPT-2 feedback encoding (ablation) | LaGEA · MT10 five observable tasks | average success rate (%): 72.67 -> 80 | Improved |
| train on one viewpoint, test on unseen overhead camera | LaGEA · MT10 five observable tasks | average success rate (%): 80 -> 79.33 | No clear effect |
| add LaGEA language-feedback reward over FuRL | LaGEA · Meta-World MT10 hidden fixed goal | average success rate (%): 76.0 -> 80.0 | Improved |
| add LaGEA language-feedback reward over FuRL | LaGEA · Robotic Fetch | average success rate (%): 44.17 -> 51.67 | Improved |
Closest papers by meaning
- AHA: A Vision-Language-Model for Detecting and Reasoning Over Failures in Robotic Manipulation (ICLR 2025 · similarity 0.90)
- ManiTaskGen: A Comprehensive Task Generator for Benchmarking and Improving Vision-Language Agents on Embodied Decision-Making (NeurIPS 2026 · similarity 0.90)
- CaP-X: A Framework for Benchmarking and Improving Coding Agents for Robot Manipulation (ICML 2026 · similarity 0.89)
- Multi-Modal Grounded Planning and Efficient Replanning for Learning Embodied Agents with a Few Examples (AAAI 2025 · similarity 0.89)
- Robot-R1: Reinforcement Learning for Enhanced Embodied Reasoning in Robotics (NeurIPS 2025 · similarity 0.89)
- Learning with Language-Guided State Abstractions (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.
