Research map / Multi-agent systems and games / LLM agents
GUIOdyssey: A Comprehensive Dataset for Cross-App GUI Navigation on Mobile Devices
Quanfeng Lu, Wenqi Shao, Zitao Liu, Lingxiao Du, Fanqing Meng, Boxuan Li, Botong Chen, Siyuan Huang, Kaipeng Zhang, 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 historical actions (ablation) | OdysseyAgent · GUIOdyssey 4-split overall (HL) | AMS: 55.60 -> 63.44 | Improved |
| add historical screenshots (ablation) | OdysseyAgent · GUIOdyssey 4-split overall (HL) | AMS: 55.60 -> 60.30 | Improved |
| add contextual information as history (ablation) | OdysseyAgent · GUIOdyssey 4-split overall (HL) | AMS: 55.60 -> 64.77 | Improved |
| add contextual information to actions + screenshots (ablation) | OdysseyAgent · GUIOdyssey 4-split overall (HL) | AMS: 63.60 -> 66.84 | Improved |
| add semantic annotations in training (ablation) | OdysseyAgent · GUIOdyssey 4-split overall (HL) | AMS: 63.60 -> 66.74 | Improved |
| add decision rationale alone as semantic annotation (ablation) | OdysseyAgent · GUIOdyssey 4-split overall (HL) | AMS: 63.60 -> 66.23 | Improved |
| add screen description alone as semantic annotation (ablation) | OdysseyAgent · GUIOdyssey 4-split overall (HL) | AMS: 63.60 -> 64.64 | No clear effect |
| add history resampler to Qwen-VL (ablation) | Qwen-VL · GUIOdyssey Test-Random (HL) | AMS: 74.67 -> 75.79 | Improved |
| use history resampler instead of multi-image concatenation (ablation) | OdysseyAgent · GUIOdyssey 4-split (HL) | AMS: 65.04 -> 63.60 | Mixed |
| add OmniParser to zero-shot GPT-4o | GPT-4o · GUIOdyssey Test-Random (HL) | AMS: 13.19 -> 23.67 | Improved |
| add OmniParser to zero-shot InternVL2-Pro (high-level) | InternVL2-Pro · GUIOdyssey Test-Random (HL) | AMS: 16.04 -> 14.69 | Hurt |
| train on both high- and low-level instructions (ablation) | OdysseyAgent · GUIOdyssey Test-Random (test HL) | AMS: 75.79 -> 78.96 | Improved |
Closest papers by meaning
- GUI-Xplore: Empowering Generalizable GUI Agents with One Exploration (CVPR 2025 · similarity 0.89)
- Agent-SAMA: State-Aware Mobile Assistant (AAAI 2026 · similarity 0.88)
- VenusBench-Mobile: A Challenging and User-Centric Benchmark for Mobile GUI Agents with Capability Diagnostics (ICML 2026 · similarity 0.88)
- Scaling, Benchmarking, and Reasoning of Vision-Language Agents for Mobile GUI Navigation (ICML 2026 · similarity 0.87)
- TongUI: Internet-Scale Trajectories from Multimodal Web Tutorials for Generalized GUI Agents (AAAI 2026 · similarity 0.87)
- ShowUI: One Vision-Language-Action Model for GUI Visual Agent (CVPR 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.
