Ramp Metering on NYC Manhattan (out-of-domain)
81.38Travel TimeGemini-3.1-Pro
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemini-3.1-ProFramework=TrafficClaw2026.04 | 81.38 | 3.83 | |
| TrafficClaw (8B)Model size=8B2026.04 | 96.09 | 3.64 | |
| Qwen3-8BFramework=TrafficClaw2026.04 | 99.77 | 3.76 | |
| TrafficGPT2026.04 | 102.99 | 4.29 | |
| o4-miniFramework=TrafficClaw2026.04 | 103.29 | 3.8 | |
| SUMO-MCP2026.04 | 106.2 | 3.53 | |
| Qwen3-235BFramework=TrafficClaw2026.04 | 108.75 | 3.68 | |
| MiniMax-M2.5Framework=TrafficClaw2026.04 | 124.93 | 4.07 | |
| Qwen3-32BFramework=TrafficClaw2026.04 | 127.86 | 3.38 | |
| DeepSeek-V3.2Framework=TrafficClaw2026.04 | 130.65 | 3.21 | |
| AIDE2026.04 | 136.25 | 4.69 | |
| Classic Method2026.04 | 138.16 | 4.08 | |
| Kimi-K2.5Framework=TrafficClaw2026.04 | 140.5 | 3.34 |