Time Series Reasoning on RCW
77AccuracyARTIST
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ARTISTmode=+ SFT + RL2026.02 | 77 | 50 | |
| ARTISTRL=true, SFT=true2026.02 | 77 | — | |
| TimeMasterRL=true2026.02 | 76.99 | — | |
| ChatTS-14Bmode=+ SFT2026.02 | 73.89 | 33.1 | |
| ChatTS-14Bmode=Base Model2026.02 | 69.91 | 30.24 | |
| ARTISTmode=+ SFT2026.02 | 69.75 | 61.46 | |
| ARTISTSFT=true2026.02 | 69.75 | — | |
| TimeMaster2026.02 | 67.7 | — | |
| ITFormer-4Bmode=+ SFT2026.02 | 67.31 | 57.95 | |
| OpenTSLM-4Bmode=+ SFT2026.02 | 65.49 | 38.29 | |
| Llama-3 8Bstatistics=w/ statistics2026.02 | 62.83 | 45.82 | |
| VL-Time2026.02 | 57.52 | — | |
| Base Model2026.02 | 50.33 | — | |
| Random Guess2026.02 | 50 | 50 | |
| VL-Timefew-shot=true2026.02 | 48.67 | — | |
| Llama-3 8Bstatistics=w/o statistics2026.02 | 43.97 | 40.41 | |
| GPT-5statistics=w/o statistics2026.02 | 34.07 | 34.88 | |
| Qwen3-14Bstatistics=w/o statistics2026.02 | 33.19 | 27.4 | |
| GPT-5statistics=w/ statistics2026.02 | 32.74 | 31.08 | |
| Qwen3-14Bstatistics=w/ statistics2026.02 | 22.57 | 22.91 | |
| Qwen3-8Bstatistics=w/o statistics2026.02 | 0 | 0 | |
| Qwen3-8Bstatistics=w/ statistics2026.02 | 0 | 0 |