Research map / Vision-language and multimodal / Multimodal LLMs
Stand Up and Move: Benchmarking Interactive Spatial Intelligence in WalkerBench
Zhiqi Ge, Gang Yang, Ziyang Pan, Jingzhe Zhu, Yuancheng Gu, Juncheng Li, Qizhou Wang, Rui Tang, Siliang Tang, Jun Xiao, Yueting Zhuang
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 Explicit Topological Memory only (ablation) | VLM agents · WalkerBench ablation (average over 9 agents) | Total accuracy (%): 18.17 -> 27.83 | Improved |
| add Goal-Directed Perception only (ablation) | VLM agents · WalkerBench ablation (average over 9 agents) | Total accuracy (%): 18.17 -> 30.14 | Improved |
| add both ETM and GDP (full Spatial-IDE) (ablation) | VLM agents · WalkerBench ablation (average over 9 agents) | Total accuracy (%): 18.17 -> 36.66 | Improved |
| add ETM on top of GDP-only (ablation) | VLM agents · WalkerBench ablation (average over 9 agents) | Total accuracy (%): 30.14 -> 36.66 | Improved |
| add GDP on top of ETM-only (ablation) | VLM agents · WalkerBench ablation (average over 9 agents) | Total accuracy (%): 27.83 -> 36.66 | Improved |
| add ETM alone for navigation (ablation) | VLM agents · WalkerBench ablation (average over 9 agents) | Nav accuracy (%): 34.67 -> 47.14 | Improved |
| add GDP alone for navigation (ablation) | VLM agents · WalkerBench ablation (average over 9 agents) | Nav accuracy (%): 34.67 -> 39.00 | Improved |
| add GDP alone for visual localisation (ablation) | VLM agents · WalkerBench ablation (average over 9 agents) | Vis accuracy (%): 11.43 -> 26.57 | Improved |
| add Spatial-IDE to claude-opus-4-6 | claude-opus-4-6 · WalkerBench | Total accuracy (%): 24.49 -> 43.17 | Improved |
| add Spatial-IDE to glm-4.6v | glm-4.6v · WalkerBench | Total accuracy (%): 13.31 -> 31.77 | Improved |
| add Spatial-IDE to qwen3-vl-8b-instruct | qwen3-vl-8b-instruct · WalkerBench | Total accuracy (%): 14.66 -> 32.17 | Improved |
| add Spatial-IDE to qwen3-vl-235b-instruct | qwen3-vl-235b-instruct · WalkerBench | Total accuracy (%): 17.37 -> 35.91 | Improved |
Closest papers by meaning
- EvoEmpirBench: Dynamic Spatial Reasoning with Agent-ExpVer (AAAI 2026 · similarity 0.89)
- SpatialClaw: Rethinking Action Interface for Agentic Spatial Reasoning (NeurIPS 2026 · similarity 0.88)
- From Where Things Are to What They Are For: Benchmarking Spatial-Functional Intelligence in Multimodal LLMs (CVPR 2026 · similarity 0.88)
- From Indoor to Open World: Revealing the Spatial Reasoning Gap in MLLMs (CVPR 2026 · similarity 0.88)
- Unfolding Spatial Cognition: Evaluating Multimodal Models on Visual Simulations (ICLR 2026 · similarity 0.88)
- SpatialScore: Towards Comprehensive Evaluation for Spatial Intelligence (CVPR 2026 · similarity 0.88)
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.
