Temporal Fact Extraction on HyperRED-Temporal (test)
68.61PrecisionTSDRE w Flan-T5
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TSDRE w Flan-T5Foundation model type=LLM, Base model=Flan-T5, Enhancement=TSD (Timeline-based Sentence Decomposition)2024.05 | 68.61 | 64.91 | 66.71 | |
| Flan-T5Model size=Large2024.05 | 66.64 | 61.06 | 63.73 | |
| Flan-T5 + ExplanationModel size=Large, Enhancement=ChatGPT3.5-generated explanation2024.05 | 64.99 | 58.85 | 61.76 | |
| Llama2 w LoRAFine-tuned=True, Adapter=LoRA2024.05 | 56.95 | 31.87 | 40.87 | |
| Llama2 w LoRA + TSDFoundation model type=LLM, Fine-tuned=True, Adapter=LoRA, Enhancement=TSD (Timeline-based Sentence Decomposition)2024.05 | 56.6 | 48.06 | 51.98 | |
| CubeRE2024.05 | 55.31 | 49.64 | 52.33 | |
| ChatGPT3.5Foundation model type=LLM, Learning mode=In-context Learning2024.05 | 11.26 | 20.08 | 14.43 |