# DyGnROLE: Asymmetric Pretraining for Edge Classification on Dynamic Graphs

Source: [DyGnROLE: Asymmetric Pretraining for Edge Classification on Dynamic Graphs](<https://www.sota2.com/research/paper/69a125162d603e6de3e3c598>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-07-01T05:00:04.992Z

Authors: Tyler Bonnet, Marek Rei

Published: 2026-02-26T15:51:51.000Z

## Abstract

Edge classification on directed dynamic graphs requires modeling interactions between source and destination nodes exhibiting asymmetrical behavioral patterns and temporal dynamics. However, existing dynamic graph architectures largely rely on shared parameters for processing source and destination nodes, with limited or no systematic role-aware modeling. We propose DyGnROLE (Dynamic Graph Node-Role-Oriented Latent Encoding), a Transformer-based architecture that disentangles source and destination representations. By using separate embedding tables and role-semantic positional encodings, the model captures the distinct structural and temporal contexts unique to each role. Critical in limited-label settings, which are common in edge classification, is a self-supervised pretraining objective we introduce: Directional Role Alignment (DRA). DRA learns distinct but aligned source and destination embedding spaces by training source representations to retrieve their corresponding destination representations while a historical positive masking strategy excludes previously observed interactions from future negative comparisons. The masks introduce a temporally directional training signal in which node pairs progress monotonically from unseen to observed, after which the relationship is eligible only for further alignment. A comprehensive evaluation on four edge classification tasks across eight datasets demonstrates that DyGnROLE consistently outperforms a wide range of state-of-the-art baselines, highlighting the importance of role-aware representation learning and asymmetric pretraining for modeling complex directed interactions when labeled data is limited.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Edge Classification on Enron](<https://www.sota2.com/research/sota/edge-classification-on-enron.md>): Edge classification; ENRON; MF1: 10.3.

- [Edge Classification on GDELT](<https://www.sota2.com/research/sota/edge-classification-on-gdelt.md>): Edge classification; GDELT; MF1: 60.

- [Post Usefulness Prediction on Stack E](<https://www.sota2.com/research/sota/post-usefulness-prediction-on-stack-e.md>): Post Usefulness Prediction; Stack E; MF1: 55.6.

- [Post Usefulness Prediction on Stack U](<https://www.sota2.com/research/sota/post-usefulness-prediction-on-stack-u.md>): Post Usefulness Prediction; Stack U; MF1: 53.9.

- [Rating Prediction on Amazon (MF1, WF1)](<https://www.sota2.com/research/sota/rating-prediction-on-amazon-mf1-wf1.md>): Rating Prediction; AMAZON; MF1 Score: 23.

- [Rating Prediction on Google (MF1, WF1)](<https://www.sota2.com/research/sota/rating-prediction-on-google-mf1-wf1.md>): Rating Prediction; Google; MF1: 22.4.

- [Edge Classification on ICEWS](<https://www.sota2.com/research/sota/edge-classification-on-icews.md>): Edge classification; ICEWS; MF1: 2.1.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2602.23135>)

[PDF](<https://arxiv.org/pdf/2602.23135.pdf>)
