arXiv:2606.01873v1 Announce Type: new 
Abstract: LLM-as-Aligner has emerged as a prevalent pre-training paradigm for Text-Attributed Graphs(TAGS), aligning graph and text modalities into a shared embedding space via CLIP-style contrastive learning. While effective on individual downstream tasks, we observe severe catastrophic forgetting when such models are sequentially fine-tuned on streaming tasks. Although parameter-efficient fine-tuning alleviates forgetting to some extent, it remains insufficient to resolve task interference and ineffective knowledge transfer. In this work, we study graph continual learning for LLM-as-Aligner models on TAGs, with the goal of mitigating interference while promoting positive transfer across tasks. This setting introduces two fundamental challenges: (1) heterogeneous downstream tasks induce shifting optimization objectives, hindering unified fine-tuning; and (2) graph and text encoders exhibit different sensitivities to adaptation, making uncoordinated updates prone to misalignment. To address these challenges, we propose G2LoRA, a continual learning framework for TAGs. G2LoRA unifies node-, link-, and graph-level tasks under a single graph--text alignment objective, and enables consistent optimization across domain/class/task incremental modes. To reduce task interference while encouraging positive transfer, G2LoRA performs category-aware gradient projection in structured subspaces, resolving conflicting updates and enabling conditional backward transfer to balance forward and backward knowledge flow. To further prevent cross-modal drift, G2LoRA introduces gradient magnitude modulation to coordinate update rates between graph and text encoders. Extensive experiments on benchmark datasets demonstrate that G2LoRA consistently outperforms strong baselines across different backbone architectures, achieving superior continual performance and transferability.

تم تقديم إطار G2LoRA لتحسين التعلم المستمر للرسوم البيانية، وخاصة للرسوم البيانية المعنونة بالنص (TAGs)، من خلال معالجة مشكلات النسيان الكارثي وتداخل المهام في نماذج LLM-as-Aligner. يهدف هذا الإطار إلى تحسين توافق أنماط الرسوم البيانية والنصوص من خلال عملية تكيف أكثر كفاءة.

Se ha presentado el marco G2LoRA para mejorar el aprendizaje continuo de grafos, específicamente para Grafos Atribuidos por Texto (TAGs), abordando problemas de olvido catastrófico e interferencia de tareas en modelos LLM-as-Aligner. Este marco busca mejorar la alineación de modalidades de grafos y texto a través de un proceso de adaptación más eficiente.

Le cadre G2LoRA a été introduit pour améliorer l'apprentissage continu des graphes, en particulier pour les Graphes Attribués par Texte (TAGs), en abordant les problèmes d'oubli catastrophique et d'interférence des tâches dans les modèles LLM-as-Aligner. Ce cadre vise à améliorer l'alignement des modalités de graphes et de texte grâce à un processus d'adaptation plus efficace.

The G2LoRA framework has been introduced to enhance graph continual learning specifically for Text-Attributed Graphs (TAGs) by addressing issues of catastrophic forgetting and task interference in LLM-as-Aligner models. This framework aims to improve the alignment of graph and text modalities through a more efficient adaptation process.