Instance Discrimination for Link Prediction

arXiv cs.LG Papers

Summary

This paper adapts instance discrimination self-supervised learning to link prediction in graphs, proposing new models L-GRACE and L-BGRL that operate on link representations and improve performance especially on unattributed graphs.

arXiv:2605.20257v1 Announce Type: new Abstract: Recently, instance discrimination models have emerged as a major solution for self-supervised learning. Having already demonstrated its effectiveness in the image domain, instance discrimination learning is now proving equally convincing in the graph domain, in particular for node classification. However, fewer contributions have tackled the link prediction task. In this contribution, we propose to adapt existing methods to this context. We first provide a rigorous evaluation of existing self-supervised models in the field of link prediction, showing that the main performance depends on the augmentation process (like in computer vision). We then propose a new structural augmentation based on the community structure that is relevant for link prediction. Our main contribution introduces two new models, L-GRACE and L-BGRL, based on link representations instead of node representations, which improve the performance of the existing methods, especially on unattributed graphs, and we show that they perform on par with the state of the art, both in supervised and self-supervised contexts.
Original Article
View Cached Full Text

Cached at: 05/21/26, 06:21 AM

# Instance Discrimination for Link Prediction
Source: [https://arxiv.org/abs/2605.20257](https://arxiv.org/abs/2605.20257)
Authors:[Valentin Cuzin\-Rambaud](https://arxiv.org/search/cs?searchtype=author&query=Cuzin-Rambaud,+V)\(SyCoSMA, DM2L, LIRIS, UCBL\),[Mathieu Lefort](https://arxiv.org/search/cs?searchtype=author&query=Lefort,+M)\(LIRIS, SyCoSMA, IRISA, MALT, UR\),[Rémy Cazabet](https://arxiv.org/search/cs?searchtype=author&query=Cazabet,+R)\(DM2L, LIRIS, UCBL, IXXI\)

[View PDF](https://arxiv.org/pdf/2605.20257)

> Abstract:Recently, instance discrimination models have emerged as a major solution for self\-supervised learning\. Having already demonstrated its effectiveness in the image domain, instance discrimination learning is now proving equally convincing in the graph domain, in particular for node classification\. However, fewer contributions have tackled the link prediction task\. In this contribution, we propose to adapt existing methods to this context\. We first provide a rigorous evaluation of existing self\-supervised models in the field of link prediction, showing that the main performance depends on the augmentation process \(like in computer vision\)\. We then propose a new structural augmentation based on the community structure that is relevant for link prediction\. Our main contribution introduces two new models, L\-GRACE and L\-BGRL, based on link representations instead of node representations, which improve the performance of the existing methods, especially on unattributed graphs, and we show that they perform on par with the state of the art, both in supervised and self\-supervised contexts\.

## Submission history

From: Valentin Cuzin\-Rambaud \[[view email](https://arxiv.org/show-email/0996fe63/2605.20257)\] \[via CCSD proxy\] **\[v1\]**Mon, 18 May 2026 13:27:41 UTC \(350 KB\)

Similar Articles

A Unified Perspective for Learning Graph Representations Across Multi-Level Abstractions

arXiv cs.LG

This paper proposes a unified contrastive framework for learning graph representations across multiple abstraction levels (node, proximity, cluster, graph) with a parameter-free self-weighting mechanism that adaptively assigns weights to similarity scores, outperforming state-of-the-art on downstream tasks like classification, clustering, and link prediction.

Dynamic Link Prediction with Temporally Enhanced Signed Graph Neural Networks

arXiv cs.LG

This paper proposes a modular temporal enhancement framework for signed graph neural networks that integrates historical context via a Historical Context Integration Module (HCIM) with LSTM and multi-head temporal attention, achieving consistent improvements on real-world temporal signed networks for dynamic link prediction.