Single-temporal motif recognition on LLMTM generalization (test)
99.84-tailedtriangle Acctool-augmented LLM agent
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| tool-augmented LLM agentBackbone=GPT-4o-mini, Method=Agent2025.12 | 99.8 | 5,883 | 100 | 5,932 | |
| structure-aware dispatcherBackbone=GPT-4o-mini, Method=Ours2025.12 | 88.6 | 3,705 | 84.8 | 3,405 | |
| SingleBackbone=GPT-4o-mini, Method=Single2025.12 | 85.5 | 3,978 | 86.5 | 3,715 | |
| structure-aware dispatcherBackbone=openPangu-7B, Method=Ours2025.12 | 84 | 6,713 | 79.8 | 6,206 | |
| RandomBackbone=GPT-4o-mini, Method=Random2025.12 | 75 | 3,843 | 77.8 | 3,864 | |
| GPT-4o-miniBackbone=GPT-4o-mini, Method=GPT-4o-mini2025.12 | 53 | 1,642 | 60 | 1,652 |