Agentic Person Search (Spatial Reasoning) on Track 2 Spatial
100TWSOracle
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OracleBackbone=Oracle2026.04 | 100 | 100 | 100 | 2.05 | |
| ARGOS AgentBackbone=Cl. Sonnet 42026.04 | 38.3 | 76 | 38.4 | 6.71 | |
| ARGOS AgentBackbone=GPT-5.22026.04 | 33.8 | 73.1 | 33.6 | 7.04 | |
| ARGOS AgentBackbone=GPT-4o2026.04 | 32.3 | 74.5 | 31.6 | 7.49 | |
| ARGOS AgentBackbone=GPT-5-mini2026.04 | 31.9 | 74.9 | 32.2 | 7.48 |