Research map / Multi-agent systems and games / Human-AI and goal-conditioned
See and Think: Embodied Agent in Virtual Environment
Zhonghan Zhao, Xuan Wang, Wenhao Chai, Boyi Li, Shengyu Hao, Shidong Cao, Tian Ye, Gaoang Wang
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 the vision unit (ablation) | STEVE · Tech tree: Diamond Tool | iterations: 158 (1/3 success) -> 106 ± 12 (3/3 success) | Improved |
| add the vision unit (ablation) | STEVE · Tech tree: Iron Tool | iterations: 46 ± 11 -> 15 ± 2 | Improved |
| replace STEVE-13B with GPT-4 behind the same vision unit (ablation) | STEVE · Tech tree: Diamond Tool | iterations: 106 ± 12 -> 89 ± 9 | Improved |
| replace STEVE-13B with GPT-4 behind the same vision unit (ablation) | STEVE · Tech tree: Wooden Tool | iterations: 4 ± 1 -> 6 ± 2 | Hurt |
| add visual perception to the agent versus Voyager (text only) | block search agent · Continuous block search | iterations to find 10 diamonds: 35 -> 14 | Improved |
| fine-tune Llama2 on Minecraft QA (STEVE-13B) | Llama2-13B · Minecraft knowledge QA | overall score: 6.89 -> 8.12 | Improved |
| fine-tune Llama2-7B on Minecraft QA (STEVE-7B) | Llama2-7B · Minecraft knowledge QA | overall score: 6.56 -> 7.94 | Improved |
| STEVE versus Voyager on the diamond tool | STEVE · Tech tree: Diamond Tool | iterations: 102 (1/3 success) vs 106 ± 12 (3/3 success) | No clear effect |
Closest papers by meaning
- Steve-Eye: Equipping LLM-based Embodied Agents with Visual Perception in Open Worlds (ICLR 2024 · similarity 0.92)
- ADAM: An Embodied Causal Agent in Open-World Environments (ICLR 2025 · similarity 0.89)
- OmniJARVIS: Unified Vision-Language-Action Tokenization Enables Open-World Instruction Following Agents (NeurIPS 2024 · similarity 0.89)
- MP5: A Multi-modal Open-ended Embodied System in Minecraft via Active Perception (CVPR 2024 · similarity 0.89)
- Optimus-1: Hybrid Multimodal Memory Empowered Agents Excel in Long-Horizon Tasks (NeurIPS 2024 · similarity 0.89)
- CrafterDojo: A Suite of Foundation Models for Building Open-Ended Embodied Agents in Crafter (NeurIPS 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.
