LiFTER Neuro-Symbolic Predictor for Continuous-Time Dynamic Graph Forecasting
August 10, 2026
LiFTER introduces a neuro-symbolic architecture that preserves interactions as grounded temporal facts and uses executable rules to predict future links. This method enables explicit inspection and intervention on the specific rules and entity bindings responsible for each prediction score.
HOW THIS AFFECTS YOU
●
researcherYou can achieve interpretable graph forecasting by treating temporal patterns as executable symbolic rules rather than opaque neural states.