Financial Time Series Reasoning on FinTSR-Bench (test C)
2Drawdown (Draw.)Phi-3.5-mini
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Phi-3.5-miniType=Language Model2026.05 | 2 | 0.3 | 10.5 | 7.3 | 0 | 0.5 | 1.1 | 0.8 | 1.8 | 0.8 | 2.5 | |
| TimeMQA-MistralType=TS Language Model2026.05 | 4.7 | 2.6 | 5.5 | 0.8 | 3.9 | 12.7 | 13.3 | 11.1 | 7.9 | 4.9 | 6.7 | |
| TimeMQA-QwenType=TS Language Model2026.05 | 5.1 | 1 | 3.3 | 0.3 | 1.2 | 2 | 1.1 | 1.7 | 1.4 | 1.4 | 1.8 | |
| TimeMQA-LlamaType=TS Language Model2026.05 | 10.3 | 6.3 | 5.8 | 6.9 | 16.6 | 18.6 | 15.7 | 8.7 | 9.7 | 10.2 | 10.9 | |
| Mistral-7BType=Language Model2026.05 | 10.4 | 0.3 | 1.4 | 0 | 46.1 | 51 | 1.8 | 40.2 | 28.3 | 36.9 | 21.6 | |
| Random2026.05 | 25 | 33.3 | 20 | 33.3 | 50 | 50 | 50 | 50 | 50 | 50 | 41.2 | |
| Llama-3.1-8BType=Language Model2026.05 | 57.8 | 30.2 | 76.4 | 2 | 46.4 | 50.2 | 46.3 | 45.9 | 31.5 | 35.1 | 42.2 | |
| TimeOmni-1-7BType=TS Reasoning Model2026.05 | 73 | 33.7 | 49.7 | 34.5 | 48.6 | 52.8 | 53.7 | 54.1 | 50.8 | 48.4 | 49.9 | |
| Qwen2.5-7BType=Language Model2026.05 | 73.7 | 36.6 | 77.5 | 25.8 | 51.4 | 50.4 | 55 | 44.9 | 51.3 | 47.4 | 51.4 | |
| Gemma-2-9BType=Language Model2026.05 | 77.3 | 0 | 50.7 | 30.8 | 45.4 | 30.6 | 23.2 | 20.1 | 0 | 0.2 | 27.8 | |
| Qwen2.5-7B (w/o CoT)Training=SFT on FinTSR-Bench, Chain of Thought (CoT)=false2026.05 | 77.9 | 39.1 | 72.8 | 30.3 | 50.4 | 55.2 | 61.5 | 59 | 50.3 | 49.4 | 54.6 | |
| Qwen2.5-7B (w/ CoT)Training=SFT on FinTSR-Bench, Chain of Thought (CoT)=true2026.05 | 96.5 | 45.7 | 91.6 | 38.5 | 48.7 | 51 | 56.4 | 49.8 | 45.8 | 53.6 | 57.8 | |
| FinSTaRChain of Thought (CoT)=true2026.05 | 99 | 92.7 | 99.4 | 87.9 | 53.5 | 79.5 | 72.9 | 78.8 | 53 | 64.6 | 78.1 |