Research map / Vision-language and multimodal / Robot action and navigation
AR-VRM: Imitating Human Motions for Visual Robot Manipulation with Analogical Reasoning
Dejie Yang, Zijing Zhao, Yang Liu
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 keypoint VLM pretraining on Ego4D human videos (ablation) | AR-VRM · CALVIN ABCD->D | average success rate: 60.0% -> 81.3% | Improved |
| add retrieval of human action videos during fine-tuning (ablation) | AR-VRM · CALVIN ABCD->D | average success rate: 81.3% -> 84.3% | Improved |
| add analogical reasoning map (ablation) | AR-VRM · CALVIN ABCD->D | average success rate: 84.3% -> 85.4% | Improved |
| fine-tune VLM transformer instead of freezing it (keypoint frozen) (ablation) | AR-VRM · CALVIN ABCD->D | average success rate: 78.6% -> 85.4% | Improved |
| freeze keypoint encoder and head instead of training them (VLM trainable) (ablation) | AR-VRM · CALVIN ABCD->D | average success rate: 82.2% -> 85.4% | Improved |
| replace GR-1 with AR-VRM on unseen scenes | AR-VRM · CALVIN ABC->D | average success rate: 61.2% -> 65.9% | Improved |
| replace GR-1 with AR-VRM with 10% training data | AR-VRM · CALVIN ABCD->D (10% data) | average success rate: 40.0% -> 45.6% | Improved |
| replace GR-1 with AR-VRM on full data | AR-VRM · CALVIN ABCD->D | average success rate: 84.2% -> 85.4% | Improved |
| replace GR-1 with AR-VRM on unseen instructions | AR-VRM · CALVIN ABCD->D (unseen language) | average success rate: 43.3% -> 46.6% | Improved |
| replace GR-1 with AR-VRM on unseen real-world categories | AR-VRM · Real robot transportation | success rate: 0.30 -> 0.53 | Improved |
| keypoint trainable with VLM frozen versus both frozen (ablation) | AR-VRM · CALVIN ABCD->D | average success rate: 78.6% -> 77.1% | Hurt |
Closest papers by meaning
- Latent Action Pretraining from Videos (ICLR 2025 · similarity 0.91)
- VLMimic: Vision Language Models are Visual Imitation Learner for Fine-grained Actions (NeurIPS 2024 · similarity 0.91)
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.90)
- CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025 · similarity 0.90)
- Object-Centric Prompt-Driven Vision-Language-Action Model for Robotic Manipulation (CVPR 2025 · similarity 0.89)
- OmniManip: Towards General Robotic Manipulation via Object-Centric Interaction Primitives as Spatial Constraints (CVPR 2025 · 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.
