Research map / Vision-language and multimodal / Robot action and navigation
EmbodiedGPT: Vision-Language Pre-Training via Embodied Chain of Thought
Yao Mu, Qinglong Zhang, Mengkang Hu, Wenhai Wang, Mingyu Ding, Jun Jin, Bin Wang, Jifeng Dai, Yu Qiao, Ping Luo
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 closed-loop link from planning to low-level control (ablation) | EmbodiedGPT · Franka Kitchen, 10 demos | success rate: 38.6 -> 50.8 | Improved |
| add closed-loop link (ablation) | EmbodiedGPT · Franka Kitchen, 25 demos | success rate: 47.3 -> 58.5 | Improved |
| add closed-loop link (ablation) | EmbodiedGPT · Meta-World, 10 demos | success rate: 62.7 -> 76.4 | Improved |
| add closed-loop link (ablation) | EmbodiedGPT · Meta-World, 25 demos | success rate: 64.9 -> 81.2 | Improved |
| add EgoCOT chain-of-thought training (ablation) | EmbodiedGPT · Franka Kitchen, 10 demos | success rate: 26.2 -> 50.8 | Improved |
| add EgoCOT chain-of-thought training (ablation) | EmbodiedGPT · Franka Kitchen, 25 demos | success rate: 36.4 -> 58.5 | Improved |
| add EgoCOT chain-of-thought training (ablation) | EmbodiedGPT · Meta-World, 10 demos | success rate: 55.2 -> 76.4 | Improved |
| add EgoCOT chain-of-thought training (ablation) | EmbodiedGPT · Meta-World, 25 demos | success rate: 58.7 -> 81.2 | Improved |
| increase demonstrations from 10 to 25 | EmbodiedGPT · Franka Kitchen | success rate: 50.8 -> 58.5 | Improved |
| increase demonstrations from 10 to 25 | EmbodiedGPT · Meta-World | success rate: 76.4 -> 81.2 | Improved |
| use EmbodiedGPT (7B LM) instead of LLaVA-13B | LLaVA-13B · image input user study | object recognition score: 8.5 -> 8.4 | No clear effect |
| use EmbodiedGPT instead of LLaVA-13B | LLaVA-13B · image input user study | plan executable score: 7.6 -> 8.4 | Improved |
Later papers that compared with this method
3 later accepted papers report EmbodiedGPT as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- EVLP: Learning Unified Embodied Vision-Language Planner with Reinforced Supervised Fine-Tuning (ICLR 2026 · better on 6 of 6 shared results)
- Octopus: Embodied Vision-Language Programmer from Environmental Feedback (ECCV 2024 · better on 4 of 4 shared results)
- PERIA: Perceive, Reason, Imagine, Act via Holistic Language and Vision Planning for Manipulation (NeurIPS 2024 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents (ICML 2025 · similarity 0.91)
- Robot-R1: Reinforcement Learning for Enhanced Embodied Reasoning in Robotics (NeurIPS 2025 · similarity 0.91)
- Embodied Task Planning via Graph-Informed Action Generation with Large Language Models (ICML 2026 · similarity 0.90)
- EmbodiedScan: A Holistic Multi-Modal 3D Perception Suite Towards Embodied AI (CVPR 2024 · similarity 0.90)
- RoboMP$^2$: A Robotic Multimodal Perception-Planning Framework with Multimodal Large Language Models (ICML 2024 · similarity 0.89)
- ManiTaskGen: A Comprehensive Task Generator for Benchmarking and Improving Vision-Language Agents on Embodied Decision-Making (NeurIPS 2026 · 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.
