Morphology Reasoning on Time-MMD Traffic (test)
43.88AccuracyKAIROSAGENT-4B (+ Turn-Level Reward RL)
Evaluation results
| Method | Links | |
|---|---|---|
| KAIROSAGENT-4B (+ Turn-Level Reward RL)Horizon=96, Category=Comparable-Scale Models2026.05 | 43.88 | |
| Llama-3.1-8B-InstructHorizon=96, Category=Comparable-Scale Models2026.05 | 43.62 | |
| KAIROSAGENT-4B (SFT-Only)Horizon=96, Category=Comparable-Scale Models2026.05 | 40.56 | |
| KAIROSAGENT-4B (+ Outcome-Level Reward RL)Horizon=96, Category=Comparable-Scale Models2026.05 | 38.27 | |
| DeepSeek-R1Horizon=96, Category=Advanced Models (reference only)2026.05 | 37.76 | |
| GPT-5.2Horizon=96, Category=Advanced Models (reference only)2026.05 | 34.95 | |
| DeepSeek-R1-Distill-Qwen-7BHorizon=96, Category=Comparable-Scale Models2026.05 | 14.29 |