New AI Predicts Your Next Move Without Training
This could make friend suggestions and fraud alerts faster and smarter.
StreamTab is a training-free framework for temporal link prediction: using past interactions to predict whether two nodes will interact in the future. It is built on a frozen tabular foundation model pretrained only on synthetic tables.
StreamTab describes each candidate pair by statistics of its past interactions and shows the model recent observed interactions and sampled negative pairs as labeled examples, from which the model scores new candidates. Its central design, regime conditioning, samples negative examples according to how candidates are sampled so the examples show the distinction the model must make.
As new interactions arrive, StreamTab updates the statistics and refreshes the examples while model parameters stay fixed. According to the article, on thirteen datasets against 21 temporal graph baselines, StreamTab achieves the highest average precision on twelve datasets under historical sampling and eleven under inductive sampling, and ranks first or stays within one point of the best on seven datasets under random sampling. With the same features and model, regime conditioning raises average precision over examples with random negatives by a median of 12.6 points under historical sampling.
- StreamTab predicts future interactions without needing to be trained on each specific dataset.
- It outperformed 21 existing methods on 12 out of 13 datasets.
- This could lead to faster, cheaper AI for social networks, fraud detection, and more.
Why It Matters
Faster, cheaper AI predictions could improve friend suggestions, fraud alerts, and recommendations in apps you use daily.