Entity State Tuning (EST) introduces persistent entity states to TKG forecasters, overcoming stateless methods' long-term dependency issues. It uses a closed-loop design with topology-aware perception and dual-track evolution. Achieves state-of-the-art results across benchmarks with code on GitHub.
Key Points
- 1.Persistent global entity state memory
- 2.Aligns structure and sequence signals
- 3.SOTA performance on TKG benchmarks
Impact Analysis
Enhances long-horizon TKG forecasting by preserving historical dependencies. Boosts diverse backbones, emphasizing state persistence in temporal AI models. Open-source code accelerates research adoption.
Technical Details
Topology-aware state perceiver injects priors into encoding. Unified temporal module aggregates with sequence backbones. Dual-track mechanism balances plasticity and stability in state updates.