mala-lab

mala-lab / PromptDyG

Public

Official PyTorch implementation of ''PromptDyG: Test-Time Prompt Adaptation on Dynamic Graphs''

10
0
100% credibility
Found May 19, 2026 at 10 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Shell
AI Summary

PromptDyG is an academic research tool that helps artificial intelligence models stay smart when analyzing networks that change over time—like social networks, transaction records, or communication patterns. Instead of retraining the entire model from scratch whenever new data arrives, this method makes small, quick adjustments that keep the model accurate even when the underlying structure of the data shifts. It works as a plug-and-play addition that can boost existing models without changing them.

How It Works

1
🔬 Discover the research

You come across PromptDyG, a new method for making AI models adapt to changing network data in real-time.

2
📄 Read the paper

You learn that this technique helps models stay accurate even when the structure of the data shifts over time.

3
🔧 Set up your environment

You install the provided configuration file and all the necessary tools are automatically prepared for you.

4
📊 Download test data

You run a simple script that fetches real-world network datasets from Stanford University for testing.

5
Choose your path
▶️
Run the built-in experiments

Launch the ready-made experiments and watch the model adapt to each new snapshot of data.

🔌
Add it to your own models

Plug the technique into your existing dynamic graph models to boost their performance.

6
📈 Watch the magic happen

The model automatically adjusts its internal settings as new data arrives, keeping predictions accurate.

🎉 Get improved results

Your model now handles changing data structures gracefully and delivers better predictions than before.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 10 to 10 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is PromptDyG?

PromptDyG is a PyTorch library for learning on dynamic graphs that change over time. It solves the problem of model degradation when graph structure shifts between training and deployment. The approach uses lightweight prompts that adapt at test time without retraining the backbone model, keeping your frozen graph neural network relevant as new snapshots arrive.

Why is it gaining traction?

The key innovation is test-time adaptation for dynamic graphs. Most methods lock the model after training, but real-world graphs evolve continuously. PromptDyG injects learnable prompts that adapt unsupervised as new data arrives, theoretically widening the margin between positive and negative node pairs. It works as a plug-and-play module on top of existing dynamic graph backbones like those in Roland, the Stanford framework it builds upon.

Who should use this?

This is for ML researchers and engineers working on temporal graph problems like fraud detection, recommendation systems, or social network analysis. If you're evaluating dynamic GNN architectures and hitting performance drops on live data, PromptDyG offers a lightweight adaptation strategy without full fine-tuning cycles. It suits teams running continuous evaluation on evolving graphs who need the model to stay sharp without expensive retraining.

Verdict

At 10 stars with a 1.0% credibility score, this is a fresh academic project with limited community validation. The environment spec is complete and reproducible, but documentation is minimal. If you're exploring test-time adaptation research or need a quick baseline for dynamic graph challenges, the code is accessible enough to experiment with. For production use, wait for peer validation and additional tooling.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.