Time Series Question Answering on TimeMQA
59.33Understanding ScoreTimeART
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| TimeARTType=open2026.01 | 59.33 | 62.34 | 49.12 | 57.41 | |
| Qwen3-maxType=closed2026.01 | 53.45 | 58.27 | 46.25 | 49.33 | |
| GPT-4oType=closed2026.01 | 50.86 | 69.65 | 50 | 66.66 | |
| Qwen3-8BType=open2026.01 | 43.1 | 35.2 | 36.8 | 25 | |
| Llama3 8BType=open2026.01 | 33.62 | 36.27 | 27.03 | 25.92 | |
| TimeMQA + Llama3 8BType=open2026.01 | 28.45 | 36.27 | 22.97 | 40.74 | |
| ChatTS 7BType=open2026.01 | 25 | 28.68 | 38.51 | 12.03 | |
| TimeMQA + Mistral 7BType=open2026.01 | 18.96 | 26.2 | 12.83 | 28.7 | |
| TimeMQA + Qwen2.5 7BType=open2026.01 | 10.34 | 24.41 | 13.51 | 26.85 | |
| DeepSeek-R1 7BType=open2026.01 | 7.76 | 12.69 | 16.89 | 7.41 |