Research map / Vision-language and multimodal / Robot action and navigation
ARNOLD: A Benchmark for Language-Grounded Task Learning with Continuous States in Realistic 3D Scenes
Ran Gong, Jiangyong Huang, Yizhou Zhao, Haoran Geng, Xiaofeng Gao, Qingyang Wu, Wensi Ai, Ziheng Zhou, Demetri Terzopoulos, Song-Chun Zhu, Baoxiong Jia, Siyuan Huang
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 language instruction (ablation) | PerAct · ARNOLD Test | average success rate (%): 17.67 -> 42.47 | Improved |
| add explicit state-value supervision head (ablation) | PerAct · ARNOLD Test | average success rate (%): 42.47 -> 43.36 | No clear effect |
| add explicit state-value supervision head (ablation) | PerAct · ARNOLD Novel State | average success rate (%): 4.49 -> 4.10 | No clear effect |
| add explicit state-value supervision head (ablation) | PerAct · ARNOLD Any State | average success rate (%): 16.36 -> 19.97 | Improved |
| add explicit state-value supervision head (ablation) | PerAct (MT) · ARNOLD Test | average success rate (%): 29.14 -> 31.00 | Improved |
| add explicit state-value supervision head, given first-phase ground truth (ablation) | PerAct · ARNOLD Test (first-phase ground truth) | average success rate (%): 55.24 -> 60.22 | Improved |
| add first-phase ground truth (grasp positions) (ablation) | PerAct · ARNOLD Test | average success rate (%): 42.47 -> 55.24 | Improved |
| train multi-task instead of single-task (ablation) | PerAct · ARNOLD Test | average success rate (%): 42.47 -> 29.14 | Hurt |
| replace CLIP language encoder with T5-base (ablation) | PerAct · ARNOLD OpenDrawer | success rate (%) | Improved |
| use voxel-grid PerAct instead of top-down image 6D-CLIPort | 6D-CLIPort · ARNOLD Test | average success rate (%): 0.84 -> 42.47 | Improved |
| use PerAct instead of single-view regression BC-Lang-ViT | BC-Lang · ARNOLD OpenDrawer Test | success rate (%): 5.88 -> 31.09 | Improved |
| evaluate on novel goal states instead of seen ones | PerAct · ARNOLD Novel State | average success rate (%): 42.47 -> 4.49 | Hurt |
Closest papers by meaning
- GravMAD: Grounded Spatial Value Maps Guided Action Diffusion for Generalized 3D Manipulation (ICLR 2025 · similarity 0.88)
- Grounding Language Plans in Demonstrations Through Counterfactual Perturbations (ICLR 2024 · similarity 0.88)
- PriorVLA: Prior-Preserving Adaptation for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.88)
- Learning with Language-Guided State Abstractions (ICLR 2024 · similarity 0.88)
- Exploring the Limits of Vision-Language-Action Manipulation in Cross-task Generalization (NeurIPS 2025 · similarity 0.88)
- PAC Bench: Do Foundation Models Understand Prerequisites for Executing Manipulation Policies? (NeurIPS 2025 · 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.
