Time-series editing on TSEdit
97.6Trend FidelityT2SP
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| T2SPLLM Backbone=GPT-5.4, Representation=T2SP2026.06 | 97.6 | 87.8 | 100 | 86.6 | 79.1 | 100 | 88 | 98.4 | 100 | 90.7 | 88.4 | 100 | |
| T2SPLLM Backbone=Gemini-3.1-flash-lite, Representation=T2SP2026.06 | 97.1 | 87.7 | 100 | 86.8 | 79.1 | 100 | 86.7 | 96.7 | 97.5 | 90.2 | 87.9 | 99.2 | |
| T2SPLLM Backbone=Qwen-3.5-9B, Representation=T2SP2026.06 | 93.5 | 84.7 | 96.3 | 86.6 | 79.1 | 100 | 87.4 | 99.3 | 100 | 89.2 | 87.7 | 98.8 | |
| T2SPLLM Backbone=Claude-haiku-4.5, Representation=T2SP2026.06 | 84.4 | 86.2 | 98.8 | 86.6 | 79.1 | 100 | 88 | 98.4 | 100 | 86.3 | 87.9 | 99.6 | |
| RawLLM Backbone=GPT-5.4, Representation=Raw2026.06 | 82.9 | 58.9 | 92.5 | 65.3 | 56.2 | 93.8 | 77.6 | 91.1 | 97.5 | 75.3 | 68.7 | 94.6 | |
| InstructTimetraining-based=true, training samples=60002026.06 | 59.8 | 61.6 | 100 | 63.8 | 31.1 | 100 | 50.3 | 4.4 | 100 | 58 | 32.4 | 100 | |
| RawLLM Backbone=Qwen-3.5-9B, Representation=Raw2026.06 | 56.4 | 42.4 | 72.5 | 46.9 | 51.3 | 83.8 | 59.5 | 76 | 81.2 | 54.3 | 56.6 | 79.2 | |
| RawLLM Backbone=Gemini-3.1-flash-lite, Representation=Raw2026.06 | 50.9 | 58.2 | 91.2 | 50.9 | 61.5 | 98.8 | 73.2 | 92.6 | 100 | 58.3 | 70.8 | 96.7 | |
| VisionLLM Backbone=Qwen-3.5-9B, Representation=Vision2026.06 | 47.8 | 20.4 | 76.2 | 72.7 | 40.6 | 95 | 63.9 | 7.3 | 100 | 61.5 | 22.8 | 90.4 | |
| VisionLLM Backbone=GPT-5.4, Representation=Vision2026.06 | 45.2 | 38.9 | 87.5 | 48.9 | 39.1 | 80 | 53.8 | 27 | 87.5 | 49.3 | 35 | 85 | |
| VisionLLM Backbone=Claude-haiku-4.5, Representation=Vision2026.06 | 43.8 | 19.8 | 98.8 | 53.6 | 17.5 | 100 | 46.2 | 0 | 98.8 | 47.9 | 12.4 | 99.2 | |
| RawLLM Backbone=Claude-haiku-4.5, Representation=Raw2026.06 | 37.7 | 66.8 | 96.3 | 49.5 | 65.3 | 100 | 71.8 | 90.4 | 98.8 | 53 | 74.2 | 98.3 | |
| ChatTime-7B2026.06 | 35.8 | 11.1 | 80 | 30.7 | 9.1 | 80 | 40.2 | 0 | 80 | 35.6 | 6.7 | 80 | |
| Verbal-TStraining-based=true, training samples=60002026.06 | 35.1 | 41.9 | 100 | 50.1 | 11.8 | 100 | 49.5 | 0 | 100 | 45.2 | 17.9 | 100 | |
| VisionLLM Backbone=Gemini-3.1-flash-lite, Representation=Vision2026.06 | 31.2 | 39.1 | 76.2 | 48.2 | 25.9 | 75 | 47 | 27.2 | 73.8 | 42.1 | 30.7 | 75 |